Cash is becoming more important as AI changes payments

Artificial intelligence is changing the payment system from both directions. Banks, card networks and fintech companies use machine learning to identify suspicious transactions, automate customer service, price credit and decide which payments deserve an extra security check. Criminals use the same tools to write convincing messages, clone voices, create fake identities and accelerate fraud. The result is a more capable but more opaque payment environment. Cash matters in that environment because it is a direct, offline, public form of money that does not depend on a smartphone, a data connection, a risk model or the approval of a private platform at the moment of payment.

That is not an argument against cards, instant transfers or digital wallets. They solve real problems. They make remote commerce possible, reduce the need to carry banknotes and can lower the friction of everyday payments. Slovakia, like the rest of Europe, has adopted them quickly: the National Bank of Slovakia says instant payments accounted for 27% of SEPA transfers made in the country at the end of 2024, and the share was still rising. The issue is not whether digital payments should exist. It is whether a society should allow one connected and increasingly automated channel to become the only practical way to buy food, travel, receive wages, donate, or help a neighbour.

Cash is becoming infrastructure again

For much of the past two decades, cash was discussed as an old payment habit. The standard story was simple: digital payments were modern, cash was fading, and the remaining task was to persuade the holdouts. That story is now too narrow. The more basic question is whether a country has enough independent ways to settle ordinary transactions when the digital layer is unavailable, contested or unsafe.

Infrastructure is easy to overlook when it works. A tap on a terminal looks simple because the machinery behind it is invisible: electricity, telecoms, a payment terminal, software updates, merchant acquirers, banks, card schemes, cloud services, fraud systems, identity checks, settlement arrangements and often a phone or wearable with enough battery. Each part can be well built. The system can still be interrupted when a link fails. A payment method is resilient not because it never breaks, but because people retain an alternative when it does.

Banknotes and coins are not free of infrastructure. They need printing, transport, storage, bank branches, cash machines, tills and security. They can be stolen, lost or damaged. Their strengths lie elsewhere. A note is usable without a network and settles immediately between two people who accept it. It does not require either person to open an app, pass a biometric check, accept a software update, or be placed into a risk category. Cash is a low-tech payment rail, but low-tech is not the same as low-value.

This matters more when digital systems are becoming more complex. Artificial intelligence is rarely a single machine deciding whether a payment goes through. It is part of a chain: a behavioural score flags a payment; a device fingerprint suggests an unfamiliar phone; a fraud engine orders a temporary hold; an automated assistant tells the customer to wait; a human review may come later. The system may be right. It may also be difficult for the customer to understand, challenge or reverse in the moment. Cash does not eliminate financial risk, but it removes a class of failure: a person cannot be locked out of a banknote by a model trained on their past digital behaviour.

The European policy response has started to reflect this shift. The ECB’s cash strategy states that euro cash should remain widely available, accessible and accepted as both a means of payment and a store of value. The European Commission’s 2023 proposal on the legal tender of euro cash was intended to protect acceptance and access across the euro area. The language is revealing. Cash is no longer treated only as a legacy product; it is being framed as a public option whose disappearance would reduce choice.

AI changes the economics of trust

A cash payment asks a limited set of questions. Is the note genuine? Is the amount correct? Is the seller willing to accept it? The interaction is local. It is not automatically converted into a behavioural profile, a marketing signal, a lending input or a fraud feature. There can be cameras at a shop and legal obligations around large transactions, but the act of paying with a €20 note does not normally create the same portable record as a card or wallet transaction.

Digital payment systems solve a different problem. They make trust scalable between people who do not know each other. A merchant in Bratislava can sell to a buyer in another city or country because payment networks, banks and fraud tools stand between them. That is a major achievement. It also means that trust becomes a calculated product. Someone has to decide whether an unfamiliar transaction is normal, whether a device is compromised, whether a payment pattern resembles laundering, whether an account should be blocked, and who should absorb a loss.

Artificial intelligence changes this calculation because it can process patterns at a scale no human team could match. It can compare transactions, devices, timing, merchant histories and known fraud indicators within milliseconds. The Bank for International Settlements has highlighted the use of advanced analytics to identify financial-crime patterns in real-time retail payments, while recent BIS work on payment fraud calls for further use of AI in detection processes. Those uses are sensible, especially when criminals move funds quickly through many accounts.

Yet a system that detects more risk also has more power to interrupt normal life. A false positive is not merely an analytic error. It may mean a family cannot pay at a petrol station, a self-employed worker cannot buy materials, or an older customer cannot move money after receiving a convincing scam call. The human cost depends on the alternatives available. A customer who carries some cash, has another bank account, or can use an offline-capable card may still complete an urgent purchase. A customer whose life is arranged around one app is exposed to the judgment of that app’s entire chain of suppliers.

AI makes digital payments safer in many cases, but it also increases the number of moments in which a payment is mediated by systems the payer cannot see. That is the central reason cash deserves serious treatment. It is not a technological rival to AI. It is a form of payment that remains outside the automated decision loop.

A payment is more than a transfer of value

The word “payment” hides several functions that do not always travel together. A person paying for groceries wants value to move from buyer to seller. They may also want finality, privacy, proof, consumer protection, speed, credit, rewards, a record for tax or reimbursement, or the ability to pay during a disruption. No single method maximises every feature.

Cash is strong on immediacy, final settlement between the people present, independence from network access and everyday privacy. It is weak on remote commerce, automatic records, chargebacks and recovery after theft. Cards are convenient, widely understood and often offer dispute procedures, but they depend on a network of intermediaries and create transaction data. Instant transfers can be fast and useful for person-to-person payments, but their speed can make authorised-push-payment fraud harder to unwind. Wallets can reduce friction, but the user may be tied to a device, an operating system, an account provider and a set of data rules they did not negotiate.

The arrival of AI does not erase these trade-offs. It sharpens them. A payment app with an AI assistant may be easier to use, but it may also encourage users to treat a conversational interface as a trusted financial adviser. A fraud system may stop more scams, but it may request more data, impose more authentication steps or make opaque judgments about what looks unusual. An automated finance tool may help someone manage bills, yet it also concentrates more information about that person’s habits in one place.

The right public question is not “cash or digital?” It is “which features must always remain available to every person?” A democracy normally avoids designing essential systems around a single supplier, a single communication link or a single point of failure. Water systems have reserves. Electricity networks have contingency planning. Hospitals maintain backup power. Payments need the same discipline. Choice is not a lifestyle preference when one choice is the fallback during failure.

This approach also avoids a false romance around notes and coins. Cash should not be presented as morally pure or operationally perfect. It is expensive to distribute, can facilitate some illicit activity and may be inconvenient for large or online purchases. The case for it is stronger and more practical: it is one of the few payment methods that lets ordinary people settle a small transaction without first asking permission from a connected system.

Cash is still used across the euro area

Predictions of cash’s imminent disappearance have repeatedly run ahead of the data. The latest ECB study of payment attitudes in the euro area found that cash was still the most frequently used payment method at physical points of sale in 2024, accounting for 52% of transactions. Its share has fallen from 59% in 2022, while card and mobile payments have gained ground. In value terms, cards were the largest single instrument, at 45%, with cash at 39%. The pattern is not a contradiction. People can use cards for higher-value purchases and cash for frequent, smaller ones.

The euro-area average also hides wide national differences. The same survey reported cash shares of point-of-sale transactions ranging from 22% in the Netherlands and 27% in Finland to 64% in Slovenia and 67% in Malta. Payment behaviour is shaped by merchant habits, bank infrastructure, fees, income, age, tourism, trust in institutions and the availability of alternatives. A policy that assumes all countries are moving at the same pace risks turning a local preference into an imposed national rule.

Even where consumers prefer digital methods, they often still want cash to remain available. In the ECB’s 2022 survey, a majority of respondents regarded having cash as a payment option as important or very important, despite a stated preference among many for cards and other cashless methods in shops. That is a useful distinction. People do not have to use cash every day to value the freedom to use it when it is the best option.

Cash also has a role outside the retail checkout. It remains a common way to make small person-to-person payments, give a child money, divide expenses, pay a local service provider, set aside a household budget, provide emergency support or keep a modest reserve. A payment system that measures only checkout shares may miss these functions. A note in a drawer is not evidence of irrationality. It can be a contingency resource, a way to avoid a fee, or money allocated for a specific purpose.

The better reading of the data is not that cash has “won” or “lost.” It is that digital payments are expanding without making cash irrelevant. The public uses different tools for different tasks. Policy should preserve the ability to do that rather than confuse declining share with dispensability.

Direct settlement without a digital gatekeeper

Cash has a property that is easy to describe but difficult to replicate digitally: transfer and settlement occur at the same moment. Once a seller accepts a genuine note, the seller has the money. There is no later batch process, network settlement cycle, account freeze, chargeback claim or card-scheme dispute embedded in the act itself. For everyday purchases, this finality is often a feature.

Digital systems divide payment into stages. A merchant may receive authorisation before receiving final funds. A bank may post a payment before it clears another layer. A card transaction can later be disputed. An instant transfer may arrive in seconds but still be subject to fraud controls or legal intervention. These structures are rational responses to distance, complexity and consumer protection. They are also reasons a digital payment is never merely a digital version of handing over a note.

The difference has implications for power. In a fully mediated system, the ability to buy depends on account access, device access, identity credentials, network availability and institutional compliance. A person can lose access because of a genuine security incident, a technical error, a regulatory hold, an expired document, a forgotten password, an overzealous fraud filter or a dispute between service providers. Each event may be rare. The combined possibility is not theoretical because digital payment is a chain, not a single switch.

Cash reduces the surface on which those events can act. It does not guarantee that every shop will accept it, and it cannot solve problems that require remote payment. It does mean that the payer is not automatically exposed to the same set of digital dependencies for a face-to-face transaction. For people living close to the edge of their budget, this distinction can be immediate. A blocked card is not an abstract inconvenience when it happens at a pharmacy or a bus station.

There is also a civic dimension. Public money that works in person without a private account gives citizens a practical relationship with the currency issued in their name. The ECB describes euro cash as a means of payment and a store of value, and its digital-euro work repeatedly states that any future digital euro would complement physical cash rather than replace it. That formulation recognises that physical and digital central-bank money do not perform identical social roles.

The blackout test

The question “will the payment app work?” is usually asked as a matter of convenience. The blackout test asks a harder question: what happens when it does not? A short outage can be caused by a local telecoms issue, a dead phone battery, a bank’s technical failure, a software update, a merchant terminal fault or a disruption in the wider electricity system. A more serious incident can involve a cyberattack, flood, extreme weather, civil emergency or geopolitical shock.

No payment method survives every condition. Cash machines need power and replenishment. Shops need light, staff and change. Digital systems may retain some capacity through backup networks and local storage. The practical aim is not to find a magical instrument. It is to make sure that a failure in one layer does not stop every route to payment at once.

Sweden offers a clear recent example of official thinking. In March 2025, the Riksbank said that payment-system resilience required the ability to pay offline with cards and use cash for essential goods. It linked preparedness not only to technical continuity but to the public’s ability to make payments during crisis or, in the worst case, war. The message was not that Sweden should abandon digital payments. It was that a nearly digital society needs deliberate fallback arrangements.

The Bank of England made a similar point in May 2026, arguing that resilience comes from retaining multiple ways to pay and noting that cash remains a practical alternative when mobile coverage drops, a phone runs out of battery or other systems are not working as expected. It referred to the large outages in Spain and Portugal in 2025 as reminders that disruption is not merely a hypothetical scenario.

Payment methods under the blackout test

The table is deliberately simple. It does not rank methods from best to worst. It shows that they fail differently. A household, merchant and country that relies on more than one route is less exposed than one that relies on a single route. Cash earns its place not because it replaces digital payments, but because it fails on a different set of conditions.

AI-era fraud changes the burden on consumers

The payment debate is often framed around speed. AI shifts attention toward authenticity. A fraudster no longer needs perfect language skills, a professional recording studio or an obvious foreign call centre to imitate a bank, a relative, a manager or a merchant. Generative tools can produce cleaner phishing messages, more plausible chats, synthetic voices and images that create pressure before the target has time to verify the request.

European authorities have been direct about the trend. ENISA’s 2025 threat landscape reported that AI-supported phishing had become a defining part of the threat environment, while Europol has warned that deepfakes can be used in crimes including CEO fraud and other deception. The OECD’s 2026 work on financial scams and frauds notes that scam techniques may use AI and deepfakes to exploit consumers.

Digital payments are not uniquely vulnerable to deception; cash scams exist, and counterfeiters adapt. The difference is speed and reach. A convincing digital fraud can direct a person from a social-media advert to a fake website, to a chat, to a transfer, and then move money across several accounts before the victim understands what happened. Instant payment systems deliver real benefits, but speed narrows the window for intervention after a victim has been manipulated into authorising a transfer.

This does not mean that consumers should retreat from digital banking. It means security advice has to be more realistic. A message that “looks professional” is no longer strong evidence. A voice that “sounds exactly like” a child, a colleague or a bank employee is no longer proof. A familiar-looking screen may be generated or copied. In an AI-shaped fraud environment, the safest habit is procedural: use a trusted number or official app you opened yourself, pause under pressure, and verify a request through a second channel.

Cash plays a narrow but real role here. A person paying a local merchant in notes is not being prompted to reveal a passcode, approve a push notification or instruct an irreversible transfer. Cash cannot protect someone who hands money to a con artist. It does, however, avoid many remote and account-takeover fraud paths. The lesson is not that cash is fraud-proof; it is that payment risk has become more varied, and a society needs more than one form of payment to match it.

Privacy is becoming a priced feature

Cash is often described as private. That is true in the ordinary sense that a small cash purchase does not automatically create a transaction record shared among a bank, merchant acquirer, payment network, wallet provider, analytics vendor, advertising system or data broker. Privacy is not absolute anonymity. Retailers can use cameras, authorities can investigate crime and large transactions can trigger legal duties. But the default data trail is much lighter.

Digital payments reverse the default. The payment record is part of the service. It helps reconcile accounts, prove a purchase, issue refunds, detect fraud, administer taxes and create consumer-protection rights. For many people, this is useful. A searchable history makes personal finance easier. It can establish a record for a rental deposit or a work expense. It can help a bank detect unusual activity. The concern is not that transaction data exists; it is that the volume, granularity and commercial value of such data rise as payments become more embedded in platforms.

AI makes that data more interpretable. A list of purchases can be classified into habits, routines, probable income, travel patterns, health-related spending, social connections, religious or political associations, stress events and financial vulnerability. A single payment says little. A long sequence can say much more. Machine learning is built to draw such inferences, sometimes accurately and sometimes unfairly.

That creates a distinction between privacy as secrecy and privacy as bargaining power. Someone may have nothing to hide yet still prefer not to generate a permanent profile every time they buy a newspaper, give a gift, buy medicine or pay a neighbour. Cash lets people retain that choice for lawful routine transactions. It is a limit on routine data extraction, not a shield against legitimate investigation.

European policy is grappling with this tension in the debate over the digital euro. The ECB says the proposed digital euro would offer a digital form of central-bank money and is designed to complement cash. The project includes discussion of privacy and offline functionality, precisely because digital public money will be judged by more than speed. Still, no digital system can reproduce the data-minimising character of physical notes in every circumstance. Hardware, updates, fraud controls and legal compliance all require some architecture.

Cash is the baseline privacy option in daily commerce. It does not oblige everyone to use it. It gives everyone a way to avoid treating every small payment as a data event.

Inclusion is more than owning a smartphone

A person can appear financially included and still be vulnerable in a cashless environment. They may have a bank account but no reliable smartphone. They may have a phone but limited data, an old device, poor signal or a cracked screen. They may share a phone within a household. They may have an account but struggle with app updates, two-factor authentication, passwords, biometric checks or the fear of making an irreversible mistake.

Digital design often assumes a confident, connected and continuously authenticated user. Many people do not fit that model all of the time. Older adults, people with cognitive impairments, people with disabilities, migrants, homeless people, residents of rural areas, victims of financial abuse and people fleeing domestic control can encounter specific barriers. The relevant point is not to stereotype these groups. It is to recognise that a system designed around a single user profile will leave someone out.

The World Bank’s Global Findex 2025 illustrates the need to look beyond account ownership. Its 2025 edition covers access to financial services alongside mobile-phone ownership, internet use and digital safety, and it identifies gaps in access to digital and financial services among women and poor adults. A bank account is a necessary tool for many modern transactions; it is not proof that a person can reliably use every digital channel at every moment.

Cash provides a lower-barrier option. It does not require a credit history, a device, a subscription, a data plan, an identity-verification journey or a successful interaction with a chatbot. It can be used by a child with parental guidance, an older person who prefers familiar routines, a visitor whose app does not work locally, or anyone caught between phone replacement and account recovery. That does not make cash the only inclusive method; accessible cards, basic payment accounts, in-person banking and well-designed public digital services are also vital. It means the removal of cash creates exclusion that other reforms must actively repair.

There is also a question of dignity. Needing help to pay is not a minor inconvenience. A person who cannot complete a transaction independently can lose privacy, autonomy and confidence. Public policy should not treat dependence on a relative’s phone, a shop assistant’s patience or a call-centre queue as an acceptable cost of progress. A payment system is inclusive when people can use it without needing to explain themselves or ask someone else to authenticate their money.

A household budget needs friction

The most overlooked argument for cash is behavioural rather than technical. Notes and coins are visible. When money leaves a wallet, the remaining amount is obvious. That physical friction can be useful for people who are trying to control spending, manage a fixed income, avoid overdrafts or separate money for rent, food and transport.

Digital payments remove friction by design. A saved card, a one-click checkout, a subscription renewal and a contactless tap make spending easy. Ease is often good. It reduces queues, helps people during travel, and makes online commerce possible. But ease also reduces the small pauses that sometimes prevent an unnecessary purchase. Apps can show spending charts, set limits and send alerts, yet those features depend on a person opening the app and accepting its categories. Cash uses a simpler form of feedback: the money is either still there or it is not.

This matters in an AI-driven retail environment. Recommendation engines, personalised discounts, dynamic promotions and conversational shopping tools are built to reduce the distance between desire and purchase. They may recognise that a customer is likely to respond to urgency, scarcity language, a small monthly instalment or a tailored offer. Retailers have used persuasion long before AI. The difference is the potential for continuous, individualised adjustment.

Cash is not a cure for compulsive spending, and it is not practical for every expense. It does give consumers a way to establish a boundary outside the commercial data and recommendation loop. Someone who withdraws a weekly amount for discretionary spending creates a hard limit that cannot be nudged upward by a push notification. Someone who pays a child’s allowance in cash teaches a concrete relationship between money, choice and depletion that an abstract number in an app may not convey as clearly.

For policymakers, the point is modest but relevant. Payment design affects financial wellbeing. A system that makes every payment frictionless may suit commerce while making budgeting harder for some households. Cash preserves a voluntary form of friction at a time when AI is being used to remove friction from almost every consumer decision.

Merchants need reliability as well as convenience

Small businesses often have a more complicated view of cash than public debate suggests. Digital payments can speed up queues, reduce cash-handling work, produce better records and make it easier to sell online. They also bring merchant fees, contract terms, hardware costs, chargeback exposure, settlement timing and dependence on telecoms and service providers. A terminal failure during a busy period can turn a convenient system into a lost-sales problem within minutes.

Cash has costs too: counting, reconciliation, deposit trips, theft risk, insurance, change management and security. The correct comparison is not “cash costs money, cards are free.” It is a comparison of different cost structures. Digital payments externalise some costs into fees and subscriptions; cash makes operational work visible. The balance varies by business type, average transaction value, location, customer mix and local banking infrastructure.

AI complicates the merchant calculation. Fraud monitoring can reduce losses from stolen cards and suspicious orders. Automated dispute management may reduce staff work. At the same time, fraudsters can create more convincing fake orders, synthetic identities and social-engineering attacks against employees. A small merchant may have little ability to inspect the models used by its payment provider or challenge a false fraud decision that blocks a legitimate customer. It may simply receive a notification that a payment was declined.

Maintaining cash acceptance gives retailers a practical fallback. It also prevents a merchant from becoming wholly dependent on one acquiring relationship, network or device ecosystem. That does not mean every business must be forced into an identical rule regardless of security, location or transaction type. Online commerce cannot take physical notes. Unattended machines have practical limits. But physical businesses serving the public should not treat cash refusal as a harmless default when it turns customers away from the country’s public money.

The European Commission’s legal-tender proposal explicitly addresses unilateral exclusions of cash and access to notes and coins. Its purpose is not to prohibit every exception. It is to make acceptance and accessibility real rather than merely theoretical.

Legal tender needs practical meaning

The phrase “legal tender” is often used loosely. It does not automatically mean that every vendor must accept every stack of notes in every setting, with no exceptions. Rules can account for distance sales, legitimate security concerns, a lack of change, agreed payment methods and other circumstances. The important question is whether the legal status of cash has a practical effect in ordinary life.

If a person can technically use cash but cannot find a cash machine, cannot deposit notes, cannot pay a public fee, cannot buy a ticket, or is routinely refused at shops, formal legal tender becomes weak. Access and acceptance are linked. A country can issue excellent banknotes while allowing the surrounding infrastructure to shrink until the notes are useful only in limited corners of the economy.

The European Commission’s June 2023 proposal sought detailed rules on the scope and effects of legal tender, as well as access to euro banknotes and coins. It specifically addressed the problem of advance, unilateral cash exclusions. That policy direction reflects a simple democratic premise: a public currency should remain usable by the public, not merely recognised in legal theory.

The legislation is also relevant to AI because automated commerce makes blanket exclusion easier. A retailer can configure self-service machines, app-based ordering or unattended checkouts around one payment channel. A city can move parking, tickets and public services into phone-first interfaces. Each decision may look efficient on its own. Across an economy, they can create a rule that participation requires a device, an account and a successful digital identity. Cash acceptance rules put a brake on that outcome where face-to-face commerce is concerned.

A healthy policy does not set cash and digital money against each other as enemies. It protects the ability to choose. The ECB’s current digital-euro work makes the same point: a digital euro, if adopted under the relevant legislation, is intended to complement rather than replace cash, and euro cash is meant to remain available and accepted. As of the ECB’s 2026 progress information, the project’s possible issuance was linked to the EU legislative process, with 2029 mentioned as a potential date if lawmakers adopt the regulation during 2026.

Slovakia’s payment transition should not become a single-track system

Slovakia is an instructive case because it combines strong adoption of modern payment tools with the institutions of the euro area. Contactless cards, mobile payments and instant transfers are familiar to many consumers. The National Bank of Slovakia says the share of instant payments among SEPA transfers reached 27% by the end of 2024. That is a real change in everyday financial habits.

The country also uses euro cash that is backed by the Eurosystem’s commitment to access and acceptance. Slovak consumers do not need to choose between being “pro-cash” and “pro-innovation.” They already live in a mixed environment. The sensible aim is to keep it mixed: digital tools for speed and remote commerce, cash for direct settlement and independence, cards and offline capability for contingency, and public rules that prevent access from becoming dependent on a narrow set of private platforms.

Slovakia’s own experience with counterfeit currency should keep the discussion factual rather than sentimental. The National Bank of Slovakia reported that 2,812 counterfeit euro coins were recovered in Slovakia in 2025, 9.7% fewer than in 2024. Counterfeiting is a real risk, but it is measurable, prosecuted and managed through public design, inspection and law enforcement. The presence of counterfeits is not a reason to discard cash; it is a reason to maintain public confidence in it.

The broader risk is concentration. A nation becomes more exposed when a routine transaction depends on the same few categories of provider: major banks, card networks, phone makers, mobile operators, cloud platforms, identity services and fraud models. These organisations are not inherently unreliable. Many invest heavily in security. But national payment resilience should not rest on an assumption that they will all remain available at the same time, in the same place, under every condition.

For Slovakia, the practical policy goal is payment pluralism. A person should be able to pay digitally when that is convenient and with cash when a device, system or institution is unavailable. That is a stronger model than forcing citizens to declare loyalty to one method.

[ARTICLE CONTINUES — BODY WORDS DELIVERED SO FAR: 5143]

continueThought for 2m 0s

The hidden price of cashless convenience

A cashless purchase can look cheaper because the cost is not handed over at the counter. The customer taps a card, a wallet token or a phone. The merchant receives a settlement record later. Somewhere in the chain, however, someone pays for network capacity, compliance systems, software updates, fraud prevention, cloud storage, customer support, device replacement, dispute handling and data protection. The payment is quick because the infrastructure is already there, not because it costs nothing.

Cash makes some costs more visible. A shop must hold change, count the till, arrange deposits and protect staff. Banks must maintain cash machines, branches or other distribution points. Central banks must print and distribute notes and replace damaged currency. Those are real expenses, and they should not be romanticised away. Yet the comparison becomes misleading when the entire cost of card and wallet payments is treated as a minor merchant fee. The cashless system also has fixed costs, common dependencies and public-security demands that become obvious only when something fails.

Artificial intelligence is increasing those hidden requirements. Fraud models must be trained, tested, monitored and refreshed as criminal behaviour changes. Customer-support systems need escalation paths when a bot cannot resolve a frozen-payment problem. Security teams need to defend payment providers from data theft, credential attacks and manipulation of automated decisions. A consumer may experience this as an extra verification step. A bank experiences it as a permanent arms race.

A cashless payment system is not costless. It shifts costs from the purse to networks, contracts, data and institutions. Cash remains a form of redundancy whose value does not appear neatly on a quarterly merchant-services invoice.

This distinction matters for public policy. If cash infrastructure is assessed only by its direct operating cost, it will often look inefficient beside a payment application that relies on systems built for many other purposes. Resilience should be valued differently. A fire extinguisher is rarely used, but it is not kept because it wins a daily efficiency contest. Cash distribution, acceptance and public familiarity have a similar quality. They preserve capacity that can be used when other arrangements are under strain.

AI agents will make payment authority harder to read

The next payment debate will not be limited to phones and cards. It will concern software agents that search, compare, negotiate, order and pay on a person’s behalf. A consumer might ask an assistant to buy the cheapest train ticket, reorder household supplies or pay a utility bill under a set limit. A business might give an agent authority to purchase routine items, reconcile invoices or release payments once a contract condition is met.

Those arrangements may save time. They also create a new question: who, exactly, gave consent to a payment? A person may set a broad instruction, but an agent may interpret it through a model, use information from a merchant site, respond to a malicious prompt or select an option that the user did not foresee. The difference between “find the cheapest” and “buy this now” becomes financially important when software acts in real time.

Payment law and consumer protection have long dealt with unauthorised transactions, mistaken transfers and card disputes. Agentic systems will blur categories. A merchant may argue that the agent followed the customer’s instructions. The customer may say that the agent was manipulated or exceeded its authority. A bank may see a technically valid authentication event but an economically irrational purchase. Automated fraud tools may face the same ambiguity: the payment may look familiar because the agent made similar purchases before, even though the customer never intended this one.

The Bank for International Settlements has noted that AI can automate payment tasks, fraud detection and compliance work, while central-bank research also points to governance and risk questions created by increasingly complex financial technologies. The technical direction is clear enough; the rules governing responsibility are not. Banks, wallet providers and merchants will need limits, confirmations, spending caps, revocation tools and audit trails that ordinary people can understand.

Cash retains a simpler authority model. The person holding the note decides whether to hand it over. There is no delegated software instruction to misread, no login token to compromise and no later dispute over whether a digital assistant had implied consent. Cash will not suit every future purchase. It remains useful as a counterweight as payments become more automated.

A fraud model needs a human exit

Fraud prevention is one of the strongest arguments for AI in payments. Banks handle a volume of transactions that no human review team could inspect manually. Systems can identify unusual combinations of location, device, timing, merchant category and transfer behaviour within fractions of a second. They can block a stolen card, pause a suspicious payment or request extra verification before money leaves an account.

The benefit is real. The failure mode is real too. A model that is too cautious can block people whose lives do not resemble its training data: a person travelling after a family emergency, a new migrant sending money home, a worker buying unusual equipment, a customer changing phones after theft, or an older person who begins using digital banking after years of relying on cash. “Unusual” is not the same as fraudulent.

A customer facing a declined purchase rarely sees the model’s reasons. They see a short message, a chatbot or a phone queue. In the best cases, the interruption prevents a loss. In the worst cases, the person is left without access to their own funds while a system designed to protect them waits for a review. The harm is greater when every alternative is also digital and connected to the same identity or account.

Financial institutions should therefore measure more than fraud losses prevented. They should measure the time needed to restore access after a false positive, the quality of explanations offered to customers, the share of cases resolved without a call centre, and whether vulnerable customers can reach a trained human quickly. These are not decorative service standards. They determine whether automated security preserves trust or turns normal life into a sequence of permission requests.

The BIS has identified governance, explainability, data quality and concentration as serious questions for AI in finance. A payment system that uses AI responsibly needs a human exit: a clear route from an automated refusal to a rapid, accountable decision. Cash is not that human exit, but it can give a person the ability to buy essentials while the digital dispute is resolved.

The danger of false confidence in biometric payments

Biometrics are often presented as the answer to the password problem. A face, fingerprint or voice appears harder to steal than a string of characters. For many users, fingerprint or face authentication is genuinely convenient. It reduces the burden of remembering passwords and can prevent casual misuse of a lost device.

The AI era changes the security equation. Face images can be spoofed in some contexts. Voices can be cloned. Video can be generated or manipulated. Device-level biometrics remain stronger than a simple phone call asking someone to “confirm” their identity, but their reliability rests on the design of the full system: liveness detection, secure hardware, fallback procedures, recovery processes and human judgment when a match is uncertain.

Biometric data has another property that passwords do not. A password can be changed after a breach. A face and fingerprint cannot be replaced. That does not mean biometrics should be rejected. It means institutions should avoid treating them as a permanent answer to identity. They are one factor in a security design, not a licence to remove every other option.

The broader risk is cultural. When people become used to accepting a face scan, voice command or app notification as proof, criminals gain new openings. A convincing voice message from a “bank” can persuade someone to reveal a code. A fake video call can create urgency. A deepfake does not have to defeat a bank’s biometric system directly; it may only need to convince a person to approve a payment themselves.

European authorities have warned that deepfakes and AI-supported phishing can strengthen impersonation scams and social engineering. Cash remains relevant because it lowers the frequency with which ordinary retail payments become identity-verification events. A person can pay a baker with a note without proving a face, a voice, a device profile or a behavioural pattern. Keeping that option matters when identity itself is becoming easier to imitate.

Payment data is becoming training material

Data is often described as the fuel of AI. In payments, the phrase should prompt caution. Transaction information is not neutral fuel. It reflects where people go, what they buy, when they work, which services they use, what emergencies they experience and which relationships they maintain. Even when individual records are protected, large datasets can produce sensitive inferences when combined with other information.

Banks and payment companies already use data for legitimate reasons. They must comply with anti-money-laundering rules, investigate fraud, identify operational problems and meet legal reporting obligations. Customers also expect useful services: automatic categorisation, suspicious-transaction alerts and a searchable payment history. The presence of these benefits does not settle the question of scope. Data collected for security should not quietly become data used to decide marketing eligibility, creditworthiness or the terms of a service without clear limits.

AI heightens the pressure to reuse information. A model may gain accuracy from richer histories and more linked data. That creates an incentive to retain, combine and analyse more than is required for the payment itself. The policy challenge is to keep purpose limitation meaningful: data should be collected and used for a stated reason, not held indefinitely because it might prove commercially useful later.

This is not only a privacy argument. It is also an accuracy argument. A model trained on payment patterns may draw conclusions that are wrong, stale or socially biased. A sudden change in spending could signal a fraud event, an illness, a new job, a separation, a bereavement or nothing important at all. Automated systems may make reasonable predictions in aggregate while making damaging mistakes about individual lives.

The ECB’s work on a possible digital euro has put privacy and offline use at the centre of public discussion. It says offline payments could offer cash-like privacy, with payment details known only to the payer and recipient. That is an important design ambition. Physical cash still sets the more direct reference point. It lets people complete lawful everyday transactions without automatically producing a dataset that can be mined, scored or repurposed.

Concentration creates a common point of failure

The public often thinks of digital payments as a competitive market because many logos appear at the checkout. Behind those logos, the system can be concentrated. A small group of card schemes, cloud providers, mobile operating systems, banks, telecoms networks, identity services and cybersecurity firms may support a large share of daily commerce. Competition at the visible edge does not always mean independence underneath.

Concentration can bring benefits. Large providers can invest heavily in security, compliance and reliability. Standardisation makes payments work across borders and reduces the burden on small merchants. The concern arises when too many services rely on the same technical or commercial foundations. A software defect, cyberattack, cloud outage or policy change can then affect institutions that appeared separate to customers.

AI may deepen this pattern. Training and running advanced models requires specialist infrastructure, expensive computing power, high-quality data and highly skilled staff. Banks may buy tools from a small group of technology vendors. Several payment providers may use similar fraud models, identity services or cloud regions. A model update that produces unexpected behaviour could then spread through multiple firms at once.

This is a systemic-risk issue, not an accusation against any single company. Financial regulators already examine third-party dependence and operational resilience because modern finance relies on outsourced technology. The European Union’s Digital Operational Resilience Act, known as DORA, created a framework for managing information and communication technology risks in financial entities, including oversight of critical third-party providers.

Cash does not solve the problem of digital concentration. It creates a separate route around it for physical transactions. A shop that can take notes and coins remains able to trade when a card network, phone wallet, cloud system or identity service is unavailable. The strength of cash is not that it is outside every institution; it is that it is outside the same institution stack.

Cyber incidents now have retail consequences

A cyberattack against a bank or payment provider can sound remote until it affects a queue at a supermarket. The practical effects arrive quickly: cards declined, cash machines unavailable, salary payments delayed, merchants unable to take electronic payments, customers receiving confusing security messages, and call centres overwhelmed. Even a contained incident can produce broad anxiety because people do not know which services are safe to use.

The International Monetary Fund reported in 2026 that cyber-enabled fraud had nearly tripled in its study period, while underreporting and data gaps made the problem harder to measure fully. The point is not that every cyber event becomes a national emergency. Modern finance combines high connectivity with rapid transaction speeds, so disruption can spread from a technical incident into everyday commerce.

Artificial intelligence serves both sides. Defenders use it to spot suspicious patterns, prioritise alerts and investigate large volumes of activity. Attackers use it to make phishing messages more credible, imitate trusted people and scale deception. The same technology that lowers the cost of detection can lower the cost of fraud. The result is not a settled advantage for either side, but a faster contest.

Public guidance often tells consumers to keep software updated, use strong passwords, activate multi-factor authentication and verify unusual requests. Those steps matter. They should not become an excuse to place the whole burden on individuals. A person who is tired, frightened or rushed is vulnerable to deception, especially when the message appears to come from a bank, employer or family member. Systems should be designed so a single mistaken click does not always result in irreversible loss.

Cash has limits in a cyber incident. It depends on cash stocks, working tills and the willingness of merchants to accept it. It is still one of the few payment methods that does not need to be patched, rebooted or connected before a local exchange can occur. Every nation planning for cyber resilience should treat access to physical cash as part of civilian continuity, not as an outdated banking service.

Cash logistics are part of national preparedness

Cash is only useful in an emergency if it is available before the emergency begins. A country cannot rely on notes locked in a distant vault while local cash machines are empty, branches are closed or merchants have no change. The logistics of cash therefore matter as much as the legal right to use it.

Distribution has become harder in many places because branch networks have contracted and ATM numbers have fallen. Those changes can be rational for individual institutions. Fewer visits to a branch mean less reason to maintain an expensive building. A cash machine with low use may not justify a private operator’s cost. The problem is that a network built solely on private demand can leave gaps in rural areas, poorer neighbourhoods, small towns and locations that become critical during disruption.

The ECB has repeatedly stated that access and acceptance are necessary for cash to function as a payment option. It also notes that declining branch and ATM networks threaten that access. The policy task is not necessarily to preserve every legacy branch. It is to set a usable standard: reasonable distance, reliable availability, transparent fees, protection for vulnerable communities and clear responsibility when commercial decisions create geographic deserts.

Cash logistics also include merchants. Shops need enough notes and coins to provide change. They need routines for storage and deposits. Transport companies, fuel stations, pharmacies, grocers and local authorities need plans for temporary digital failures. Those arrangements should be tested, not assumed. A written contingency plan that no staff member understands will not help during an outage.

Households have a role as well. Holding a modest amount of cash for essential purchases is not a prediction of catastrophe. It is basic preparation, much like keeping a torch, medicine or a charged battery pack. The appropriate amount differs by household and should never be presented as a reason to hoard. The useful principle is simple: keep enough physical money to manage a short disruption without relying on a phone, card network or cash machine.

Public services must not require a private device

A cashless rule can be particularly harsh when it appears in public services. People may be asked to pay for transport, parking, municipal fees, school meals, cultural venues or administrative charges through a mobile app or online account. Digital options can reduce queues and administrative work. They can also create a condition of access: own a compatible device, maintain a connection, understand the interface and pass the required verification.

Public authorities have a higher duty than a private retailer to avoid excluding residents. A citizen should not lose practical access to a public service because a battery died, a phone was stolen, an app is unavailable in the person’s language, or an account requires documents they cannot produce quickly. A system may need identity checks for some services, but payment should not become an unnecessary identity test.

Transport offers a clear example. A city may prefer app-only tickets because they are cheaper to issue and easier to inspect. Yet visitors, children, older passengers and people without smartphones then face a barrier that has nothing to do with their willingness to pay. Paper tickets, cash options, staffed points of sale or reloadable non-smartphone cards can preserve access. The right option depends on the service, but a public system should offer at least one route that does not require a private digital ecosystem.

The same principle applies to emergency information and assistance. During a disruption, people may need food, fuel, travel and shelter. Cash acceptance in those settings is not merely customer service. It is part of the state’s capacity to act when networks are strained. The ECB’s analysis of the 2025 Iberian blackout describes cash as indispensable when digital infrastructure fails and notes its role in reassuring the public.

A public payment channel should remain public in practice. Citizens should not have to rent access from a phone maker, mobile operator and financial platform before they can complete a basic civic transaction.

Economic abuse can begin with a blocked account

Financial control is a common feature of domestic abuse. An abusive partner may monitor spending, take control of cards, demand access to banking apps, intercept one-time codes, track a phone, empty an account or prevent a person from working. A fully digital payment environment can make that control easier when all money is visible in shared statements, all authentication routes go through a monitored device, and leaving the system requires the abuser’s cooperation.

Cash is not a complete answer. An abuser can steal it, search for it or control access to it. But it can provide a degree of immediate independence. A person may be able to retain small amounts for transport, food, a phone top-up, an emergency night away or a confidential conversation. Because cash does not automatically create a shared transaction record, it can reduce the risk that every small action becomes visible to the person exerting control.

This subject requires care. Advice to keep cash may be unsafe for someone whose belongings are searched or whose movements are monitored. Specialist domestic-abuse services are better placed to assess risk and help create a safety plan. The policy point is broader: payment design has consequences for personal autonomy that are not visible in ordinary product comparisons.

Banks have begun to recognise economic abuse as a financial-services issue, and many offer ways to change contact details, set up safer access or speak to specialist teams. Those supports are necessary. They do not remove the value of an independent payment option that is not tied to a shared account or device.

AI could make coercive control harder to detect and harder to escape. Automated monitoring tools can flag unusual spending to account holders. Location data and transaction histories can be combined. Synthetic messages can impersonate a victim or their support network. At the same time, AI could help institutions identify patterns of coercion if used carefully and with human oversight. The crucial safeguard is consent. A payment system should not make every person’s ability to leave, travel or buy essentials dependent on the person who has access to their phone or banking profile.

Cash lets children learn that money runs out

Financial education is often discussed in terms of applications, dashboards and gamified savings tools. Those tools can be useful, especially for teenagers who will manage most adult finances online. They do not replace the basic experience of handling a limited amount of money.

Cash teaches a simple rule: spending reduces what remains. A child who has €10 for a school trip can see the choices involved. Buy a snack now, keep money for later, lend a coin to a friend, or save it. The feedback is immediate. There is no hidden credit limit, no subscription renewal and no abstract balance that can be ignored until a notification arrives.

This is not an argument that cash automatically produces financial discipline. Adults can spend impulsively with notes as easily as with cards. It is an argument about learning. Physical money gives young people a concrete introduction to exchange, change, budgeting and saving before they confront credit products, buy-now-pay-later offers, personalised advertising and algorithmic shopping environments.

AI-powered commerce will make those environments more persuasive. A young consumer may receive personalised suggestions, conversational sales prompts and automated comparisons that turn a preference into a purchase in seconds. Financial education needs to include digital literacy, scepticism toward online offers and an understanding of data collection. Cash belongs in that education because it shows the tangible side of money.

Parents and schools do not need to treat digital and cash education as opposites. A child can learn to use a bank card responsibly and still be given a cash budget for a particular purpose. The useful habit is to understand that payment convenience does not eliminate scarcity. Cash makes the cost of a choice visible, which is a useful lesson when digital systems are designed to make payment disappear into the background.

The informal economy argument requires honesty

Critics of cash often point to tax evasion, undeclared work, money laundering and other illicit activity. Those concerns are legitimate. Large cash transactions can make it harder to follow the trail of funds. Criminals use whatever tools suit them, including cash, shell companies, digital assets, trade invoices and bank transfers. A serious defence of cash cannot deny its misuse.

The policy mistake is to treat lawful daily cash use as proof of suspicious intent. A person paying a plumber, buying food at a market, giving a gift or withdrawing money for a household budget is not automatically avoiding scrutiny. The vast majority of low-value cash activity is ordinary life. Eliminating that activity would impose a broad privacy and access cost on everyone while leaving determined criminals free to migrate to other methods.

Targeted regulation is more defensible than blanket exclusion. Rules can focus on high-value transactions, suspicious patterns, reporting obligations and professional gatekeepers. Law enforcement can investigate credible evidence. Banks can use advanced analytics to detect criminal networks in digital flows. The BIS’s Project Hertha explored ways of using transaction analytics to identify financial-crime patterns in real-time payment systems while applying data-minimisation principles.

Cash also does not operate outside the law. Counterfeit detection, anti-money-laundering rules, tax enforcement and criminal investigation all apply. The appropriate aim is proportionality: reduce serious abuse without making every citizen’s normal purchase traceable by default.

The same honesty is needed in the opposite direction. Digital payments are not automatically clean. They can be used for fraud, mule accounts, laundering, ransomware demands and cross-border crime. A record exists, but records only help when institutions can analyse them, exchange information lawfully and act in time. The fight against financial crime depends on targeted capability, not on the assumption that forcing all consumers into digital payments will remove wrongdoing.

Cash supports trust in public money

Most money used every day is not central-bank cash. It is commercial-bank money held in accounts and moved through private payment systems. That arrangement is familiar and useful. Bank deposits are backed by regulation, deposit-insurance arrangements and central-bank liquidity systems, but they remain claims on banks. Physical euro banknotes, by contrast, are direct central-bank money.

For many people, this distinction is invisible until a period of stress. During a banking scare, a cyber incident or a disruption of digital payments, the ability to hold and use central-bank money directly becomes psychologically and practically important. It provides a form of settlement that does not rely on the immediate operation of a commercial bank’s app, card processor or customer-service line.

The Eurosystem describes cash as central-bank money and says it remains part of people’s freedom to choose how they pay. The language should not be read as a claim that bank deposits are unsafe in normal times. It is recognition that public money performs a distinct role in the monetary system. It is universally recognisable, issued by a public institution and not dependent on a commercial contract at the point of face-to-face use.

A possible digital euro may add another form of public money for digital use. The ECB says a future digital euro would complement cash, support online and offline payments, and could be issued during 2029 if EU lawmakers adopt the relevant regulation during 2026. The project is important, but it does not remove the need for notes and coins. A digital public instrument still requires a device or card, operational rules, technical resilience and user confidence. It would expand the public-money toolkit; it would not turn physical cash into a redundant object.

Cash is a visible reminder that money is a public institution as well as a private service. That reminder becomes more valuable when the rest of the payment system is mediated by commercial platforms and increasingly influenced by AI.

Financial literacy now includes payment literacy

A person who knows how to budget but cannot tell a genuine bank message from a convincing fake is financially exposed. A person who understands compound interest but cannot recover a locked account, identify a phishing screen or find an alternative way to pay during an outage is exposed in a different way. Financial literacy needs to expand beyond household arithmetic. It now includes payment literacy: knowing which payment method fits a situation, what information a transaction creates, how fraud works, what consumer rights apply and where to go when a digital system fails.

Schools, employers, banks and public authorities often teach digital safety as a checklist. Do not share a password. Do not click unfamiliar links. Use two-factor authentication. Those are sensible habits, but the advice is no longer enough on its own. AI-generated scams do not always look unfamiliar. They can arrive in fluent Slovak, use a real company logo, mention a genuine recent event, imitate a family voice or claim that a payment must be made immediately to prevent an account freeze. The person under pressure is not necessarily careless. They are facing a more persuasive form of deception.

Payment literacy therefore needs practical exercises. People should know that a bank will not require a customer to move money to a “safe account” through an unexpected message. They should know that a caller’s voice is not proof of identity. They should know that opening a bank app independently is safer than following a link in a text. They should know the difference between a card payment, an instant transfer, a direct debit and a cash withdrawal, because each creates different possibilities for reversal and dispute.

Cash has a place in that education because it makes payment choices visible. A note is not mediated by an account screen. The act of handing it over is obvious. It has no automatic refund path and no remote delivery function, but it provides a reference point from which to understand what digital systems add: records, intermediaries, credit, protections, data and dependencies. Without that reference, consumers may mistake a payment interface for money itself.

The goal is not to turn citizens into cybersecurity specialists. It is to give them enough practical knowledge to pause before acting and enough choice to avoid being trapped by one channel. A payment-literate public knows both how to use digital money safely and why access to cash remains part of personal resilience.

Offline payment is not the same as cash

The word “offline” is used loosely in payment marketing. It can mean that a card terminal stores a transaction temporarily and sends it later. It can mean that a wallet keeps limited credentials on a device. It can mean that a payment order is created without a connection but is not completed until the connection returns. These arrangements can be useful. They should not be confused with the simple ability to exchange a note for goods.

A genuine offline digital payment has to answer difficult questions. How does a device prove that the payer has sufficient funds? How does it prevent the same balance from being spent twice? What happens if a phone is lost before it reconnects? Who bears the loss if the merchant accepts a payment that later cannot settle? How are limits set, and who can change them? Cash handles these questions through physical possession and final handover. A digital system must solve them through secure hardware, rules and risk controls.

The distinction matters during an outage. A merchant may believe a card terminal can work offline, only to discover that a particular card, terminal configuration or transaction value requires online authorisation. A customer may assume a phone wallet will function without signal, only to find that it needs a recent connection or device authentication. The failure is not necessarily a defect; it may be a feature of the service design. But people need clear expectations before they stand at a checkout with essential goods.

Sweden’s central bank has made the issue concrete. Its work on payment preparedness has included an agreement intended to expand offline card payments for essential goods, using a physical card and PIN. The Swedish example is instructive because it treats offline capability as a national resilience issue rather than a premium feature for a small group of users. It also makes clear that the capability has boundaries. It is designed for specified conditions, not as a permanent replacement for connected payment systems.

Cash remains distinct even in a future with better offline cards and offline central-bank money. It does not need a previously provisioned device, a battery, a PIN, a software patch or a transaction limit established by an issuer. It is not automatically better for every purpose. It is a different instrument with a different failure pattern. Public resilience improves when people have cash, physical cards with offline capability and digital payment methods, rather than being asked to rely on one of them alone.

The design of digital money will decide its public legitimacy

The debate over a digital euro is often described in technical language: infrastructure, settlement, intermediaries, limits, wallets and privacy controls. Those questions matter, but public legitimacy will rest on ordinary experiences. Can a person use it without a smartphone? Can they pay when a connection is weak? Can a shop accept it without unreasonable cost? Will a routine transaction create a long-lived profile? Can a person recover access after a lost device? Will cash still be accepted if digital public money becomes common?

The ECB’s stated position is clear on the last point. A possible digital euro is intended to complement physical cash, not replace it. The current timetable remains conditional on EU legislation: the ECB says that, if the regulation is adopted during 2026, the Eurosystem aims to be ready for a potential first issuance during 2029. The conditional language matters. A digital euro is not in people’s wallets today, and the final decision on issuance has not been taken.

The project nevertheless exposes a useful principle. Digital public money should not merely copy the commercial payment market. If it offers no stronger privacy, no better resilience, no practical offline option, no low-cost acceptance and no meaningful inclusion, it risks becoming another account layer rather than a public alternative. If it is designed well, it could reduce dependence on a small number of private payment schemes and give Europeans another way to use central-bank money in a digital setting.

That possibility should not lead policymakers to neglect physical cash. A digital euro could be unavailable during a wider communications failure. It may need identity checks and recovery procedures. It could be used through a card as well as a phone, but it would still rely on a technological system. Notes and coins remain the only form of euro public money that people can use face to face with no electronic mediation at the moment of exchange.

There is a deeper point about consent. A person can choose not to use a new digital service. That choice becomes hollow when shops, public bodies and employers gradually remove all other routes. The public legitimacy of digital money depends on preserving exit options: cash, human support, accessible physical tools and a clear ability to refuse non-essential data practices. A digital euro would strengthen monetary choice only if it expands the number of practical options rather than narrowing them.

Stablecoins do not remove the case for public cash

Private digital tokens backed by traditional assets are often presented as the next stage of payments. Some are designed to maintain a stable value against a currency such as the euro or dollar. They may settle on distributed ledgers, connect to online platforms or move across borders outside conventional card networks. Their advocates see faster settlement, programmable transactions and new forms of competition.

Their risks are different from the risks of physical cash. A stablecoin holder depends on the issuer’s reserves, governance, legal structure, redemption process, technology, wallet provider and sometimes the rules of a trading platform. The name “stablecoin” describes an ambition, not a guarantee. Stability depends on whether the issuer can honour redemption at par under stress and whether users retain practical access to that redemption.

Europe’s Markets in Crypto-Assets Regulation, known as MiCA, has created a regulatory framework for crypto-assets, including asset-referenced tokens and e-money tokens. That is a major step toward clearer rules. Regulation does not make a private token identical to public money, however. A regulated private issuer can still face operational disruption, commercial failure, cyberattack or a loss of trust. Consumers need to understand the difference between a token that references a currency and the currency itself.

AI could make token-based finance easier to use by allowing agents to route payments, monitor wallets, compare fees and execute coded conditions. It could also make it easier to launch fraudulent tokens, create convincing impersonation campaigns and manipulate users into granting wallet permissions they do not understand. A person who signs a malicious transaction may find that there is no bank to call and no ordinary refund process.

Cash offers a contrasting form of finality and simplicity. It is not programmable. A note cannot automatically release funds when a shipment arrives, set spending limits or travel instantly across a global network. Those limits are part of its value in everyday life. It is public money, transferable without dependence on an issuer’s app or a smart contract’s code. The growth of privately issued digital money makes the continued availability of public cash more important, not less.

Programmable payments need strict boundaries

“Programmable money” is a phrase that produces both excitement and fear. The practical distinction is often missed. A payment can be programmed without the money itself being permanently controlled. A person may set a recurring rent payment, a business may release funds when an invoice is approved, or a government may issue a time-limited voucher for a specific emergency purpose. These uses attach conditions to a transaction.

The concern begins when conditions become a general feature of ordinary money. Could a payment be blocked because it falls outside a category? Could benefits expire before a person has a realistic chance to use them? Could a private platform limit a customer’s purchases based on behavioural scores, advertising profiles or a dispute with another service? Could a government or company attach restrictions that are difficult to challenge? Technical feasibility does not automatically create democratic legitimacy.

Cash has no built-in programmability. A €50 note does not know who held it before, what it was used to buy or what a software policy says it should be used for. That makes cash less convenient for some targeted public programmes. It also protects the general-purpose nature of money. Once an ordinary payment instrument becomes conditional by default, people need legal safeguards that go beyond a privacy notice or a product setting.

The digital euro debate has brought this issue into public view. The ECB has repeatedly said that a digital euro would not be programmable money and that it could not be used to impose restrictions on where, when or to whom people can pay. Its design would allow conditional payments initiated by users, such as automated transfers, but the distinction between user-directed automation and issuer-directed control is central.

Policy should preserve that distinction across the payment market. A consumer may choose a spending cap or an automatic bill payment. A merchant may set the terms of a sale. Neither fact justifies hidden restrictions on ordinary money imposed by a platform, employer, insurer, lender or public authority. Cash is a constitutional reminder that money should remain general-purpose unless a specific, lawful and transparent rule says otherwise.

Payment choice has a competition dimension

Cash is often discussed as a consumer preference or a social inclusion issue. It also affects competition. A merchant that accepts only one digital payment route is dependent on the fees, rules and technical standards of that route. A country that relies heavily on a small group of international card schemes, mobile operating systems and cloud providers is exposed to commercial as well as operational leverage.

The European Central Bank has noted that many EU countries rely entirely on international card schemes for card transactions and that international schemes accounted for a majority of euro-area card payments in 2022. The point is not that international networks are inherently undesirable. They have enabled cross-border commerce and consumer convenience. It is that the European payment market has strategic dependencies that become more visible as cash use falls and digital payments expand.

A merchant who can take cash retains at least one payment channel that does not require a card-scheme contract at the moment of sale. A customer who carries cash retains a way to purchase from a small trader who cannot afford sophisticated payment technology or who faces a temporary terminal problem. This is not a complete answer to market power, but it reduces the pressure to accept every commercial term simply because no other payment method is viable.

Competition also matters for innovation. If all retail payments flow through a few dominant platforms, new entrants have to negotiate access to those platforms or build around them. A public digital euro may offer one route toward greater European payment autonomy. National instant-payment systems, open banking, interoperable standards and consumer-friendly offline cards may offer others. Cash remains the common fallback that requires no platform onboarding at all.

The strongest payment ecosystem is not one in which a single method defeats every rival. It is one in which customers and merchants can move between methods without losing the ability to trade. Cash constrains the power of digital gatekeepers because it remains usable even when a platform chooses different terms, changes its rules or temporarily fails.

Fees and surcharging can quietly reshape access

Payment policy can change without a law being passed. A bank may close a branch. An ATM operator may introduce a fee. A merchant may remove a cash register from a self-checkout area. A delivery platform may allow only in-app payment. A public service may make the digital channel cheaper and the cash channel inconvenient. Each decision can be defended as a business choice. Their combined effect can make cash technically legal but practically burdensome.

Fees are especially important for low-income households. A person with a large balance may absorb a cash-withdrawal charge, a monthly account fee or a mobile-data cost without much thought. Someone managing a tight budget may not. Repeated small costs can push people toward a payment method they did not choose freely, especially if the nearest free ATM disappears or a cash deposit becomes expensive.

The same issue appears on the merchant side. Card acceptance can involve fixed charges, variable fees and equipment costs. Cash handling can involve banking fees, security arrangements and time. Policymakers should not assume that one method is naturally cheaper across all businesses. Small transactions, rural settings, seasonal commerce and very low-margin shops can produce a different calculation from a large urban chain.

A good policy starts with transparency. Consumers should know where they can withdraw and deposit cash without surprise charges. Merchants should understand the full cost of card acceptance, including disputes and contract conditions. Public bodies should monitor geographic gaps rather than waiting for communities to lose access. The ECB has stressed that effective cash access must be available in both urban and rural areas, including withdrawal and deposit facilities.

A payment choice is not genuine when one option is technically available but priced, located or designed out of ordinary use. Cash policy should be judged by the real journey a person must make to obtain and spend notes, not only by the formal existence of legal tender.

The right to pay must include the right to be paid

Attention often centres on consumers: can a person pay with cash, card or transfer? Workers and small businesses also need usable ways to receive money. A fully digital wage system can be convenient and safer than carrying payroll envelopes. It can also exclude people who cannot open or maintain an account, whose identity documents are in transition, or who face debt collection, account freezes or coercive control that makes an account unsafe.

Most employers should pay wages through regulated accounts because the record protects both worker and employer, and labour law may require traceability. The principle is not that every income should be paid in cash. It is that people should have a realistic route into the financial system and should not be abandoned when digital access breaks down. A worker whose bank app is unavailable should still be able to buy lunch or travel home. A sole trader facing a payment-provider outage should have a way to take a small sale.

Instant payments make receiving money faster, but they do not remove all dependence. The sender and recipient still need functioning accounts, a connection, a payment interface and a provider that has not halted the transfer. People who are paid through platforms may also be subject to automated risk rules, identity checks and reserve policies that are difficult to contest. One model can affect access to income as well as access to spending.

Cash is particularly relevant at the margin: tips, small refunds, emergency assistance, informal support between family members and temporary work arrangements that are lawful but time-sensitive. It lets value move when accounts cannot. It should not be used to conceal unlawful employment or evade tax. The distinction between a legitimate cash payment and undeclared labour is important and enforceable.

Economic participation depends on the ability both to spend and to receive money. A resilient system keeps formal digital channels strong while preserving practical options for people who are temporarily, permanently or involuntarily outside them.

Disaster planning should name payment failure directly

Emergency plans often mention food, water, medicines, fuel, transport and communications. Payments are sometimes buried inside broader references to finance or commerce. That is a mistake. People cannot reliably obtain essentials if shops cannot process transactions, cash machines are empty, bank apps are unavailable and merchants do not know which forms of payment will work.

The 2025 Iberian blackout brought this issue into public discussion. The ECB’s analysis of the event described a sharp increase in demand for cash after the power interruption and concluded that cash’s tangible, offline and broadly accepted character becomes especially important during crises. A single event does not prove that every household should prepare for every scenario. It does show that payment failure is a real part of modern disruption, not an abstract concern reserved for financial specialists.

A responsible emergency plan should identify essential merchants and services, clarify which payment methods they can accept during a loss of connectivity, maintain adequate cash distribution, and train staff for manual or offline procedures. Local authorities should test the plan with retailers, banks, telecoms operators, transport providers and emergency services. Testing matters because a plan that assumes every terminal works, every cash machine is stocked and every employee remembers the procedure is not a plan at all.

Households need proportionate preparation. A small reserve of notes in mixed denominations may be more useful than a large note that a small merchant cannot change. People should store cash securely, review it periodically and avoid public discussion of what they hold. The purpose is not fear. It is to preserve the ability to buy necessities during a short interruption.

Artificial intelligence can support preparedness by identifying supply bottlenecks, monitoring abnormal transaction patterns and helping institutions communicate during an outage. It cannot substitute for a means of payment that works when the devices and networks around it do not. Payment continuity should be treated as a core part of civil preparedness, with cash as one of its tested tools.

The environmental case needs full accounting

Cash is sometimes dismissed as environmentally backward because notes and coins have to be produced, transported and replaced. Digital payments, by contrast, appear intangible. That contrast is too simple. Digital payments depend on phones, terminals, routers, servers, data centres, fibre networks, mobile towers, plastic cards, batteries and frequent device replacement. AI adds computing demand, especially where large models are trained or used at scale.

A credible environmental comparison needs a full life-cycle view: materials, manufacturing, energy, transport, maintenance, data transmission, hardware turnover and end-of-life disposal. It must also account for use patterns. A banknote may change hands many times over several years. A phone supports many functions beyond payment. A payment terminal may process thousands of transactions. There is no meaningful answer without defining the product, transaction type, geography and lifespan being compared.

The ECB has worked on reducing the environmental footprint of euro banknotes and says the Cash 2030 strategy includes making them sustainable as well as secure. That work is appropriate. The same scrutiny should be applied to digital payment infrastructure, including the resource cost of the AI systems increasingly used for fraud analysis, customer service and personalisation.

Environmental policy should avoid using simplified claims as a reason to eliminate choice. The priority is to reduce avoidable waste in every payment system: longer-lived devices, efficient data centres, recyclable materials, durable terminals, sensible note design and careful cash logistics. A lower-carbon economy will still need payments that work during outage, protect privacy and include people without the latest hardware.

The environmental question is not whether cash or digital payments look cleaner in a marketing image. It is which mix of systems delivers reliable public service with the lowest credible lifetime cost.

A balanced cash policy does not freeze the past

Protecting cash does not mean preserving every old banking practice unchanged. Some branches will close because people use them less. Some customers will prefer mobile wallets for nearly every purchase. Many businesses will continue to move online, where physical cash cannot be used directly. The policy goal should not be nostalgia or resistance to technology.

A balanced policy sets outcomes rather than prescribing one business model. People should be able to withdraw and deposit cash within a reasonable distance. Merchants serving the public in person should accept cash except where a clear, proportionate exception applies. Essential services should have payment-continuity plans. Digital methods should be accessible, interoperable, secure and backed by human support. Fraud decisions should be reversible in time for people to manage their lives.

This approach leaves room for innovation. Banks can improve instant payments. Card networks can strengthen offline capability. Wallet providers can reduce data collection. Public authorities can build better digital services. A digital euro may add a new public option. None of these advances requires the removal of notes and coins. In fact, the existence of cash can keep digital providers honest by giving customers a credible alternative.

The Riksbank’s recent work offers a useful model of practical balance: strengthen offline card payments for essential goods while retaining cash as part of payment preparedness. The Bank of England has made a similar case for retaining multiple ways to pay, describing cash as a practical alternative during everyday and severe disruption. These are not anti-digital positions. They are resilience positions.

The future of money will be mixed. Physical cash, commercial-bank deposits, instant transfers, payment cards, digital wallets and potentially central-bank digital money will coexist. The public interest lies in making that coexistence real rather than theoretical. The country that treats payment diversity as a strength will be better prepared for technological change, fraud, outages and the unequal realities of everyday life.

Cash remains the unautomated option

Artificial intelligence will keep spreading through finance. It will screen transactions, answer customer questions, identify scams, generate credit assessments, support compliance teams and eventually carry out limited payment tasks on behalf of users. Some of those uses will reduce losses and make financial services easier to manage. Others will create new forms of error, dependency and manipulation.

The question is not whether AI belongs in payments. It already does. The question is whether human beings retain a practical way to pay when the model is wrong, the connection is down, the device is missing, the account is frozen, the fraud system is overcautious or the digital environment has become too intrusive.

Cash answers that question for face-to-face trade. It does not need to become dominant again to remain important. It needs to be obtainable, accepted and understood. A banknote in a wallet is not a rejection of modern finance. It is a claim on a public system that works without an algorithm deciding whether the purchase is normal.

That claim carries a wider lesson. As more decisions are automated, society needs spaces where ordinary action does not depend entirely on profiling, optimisation or permission. Money is one of those spaces because it sits beneath food, travel, housing, work and personal independence. Cash protects a small but fundamental area of direct human choice in an economy where more choices are being filtered through machines.

[ARTICLE CONTINUES — BODY WORDS DELIVERED SO FAR: 13,871]

continueThought for 1m 44s

Cash access is also an access issue for migrants and visitors

A payment system becomes harder to use when it assumes that every person has a local bank account, a local phone number, current identity documents, a compatible smartphone and a long-established digital history. That assumption excludes more people than it first appears to. A visitor arriving from outside the euro area may have a card that does not work at a ticket machine. A newly arrived worker may be waiting for a residence document or a bank-account opening process. A refugee may have lost a phone, wallet, documents or all three. A student may have a foreign card with fees that make small purchases costly.

Cash is often the bridge in those situations. It allows someone to pay for food, transport, a room or a local service without first proving that they belong to a particular financial ecosystem. Notes and coins are imperfect for travel: they must be obtained, stored and exchanged, and they are inconvenient for bookings made in advance. They still provide a common form of money that does not ask a person to demonstrate digital continuity before they can take part in everyday life.

That function deserves more attention as borderless commerce becomes more digital. A tourist can usually download a local transport app, but “usually” is not a public-service standard. The app may require a phone number from a supported country, an account with a global platform, a card that passes a fraud check, or a mobile connection at the exact moment of purchase. A person who fails one of those steps should not be made to feel as though they have failed a test of modernity. They are trying to buy a bus ticket.

AI adds a new complication. Fraud and identity systems often use signals that correlate with established digital behaviour: familiar devices, known locations, transaction history, phone reputation, document quality and patterns of use. Those signals can be useful. They can also make a recent arrival look risky simply because there is little history to compare. A model cannot know the whole story of a person rebuilding a life in a new country. It sees a sparse and unfamiliar record.

Financial institutions need procedures that distinguish genuine caution from automatic exclusion. They should offer accessible account-opening routes, in-person assistance and review options when automated identity checks fail. Public services should retain payment options that do not make local digital status a condition of access. Cash remains a practical equaliser because it treats a lawful euro as usable money regardless of the holder’s device, data trail or place in an algorithmic profile.

The point is not that migrants or visitors should be pushed toward cash. Many want the same digital tools as everyone else, and access to a safe bank account is important for work, housing and independence. The point is that digital inclusion must be earned through service design, not assumed through ownership of a phone. Until every person can use a digital payment route reliably, cash remains part of the route into economic participation.

Universal design should apply to payments

Payment design is often judged by the average user moving quickly through a familiar shop. That test leaves out the people for whom ordinary interfaces are difficult: someone with limited vision, a tremor, a learning disability, reduced memory, hearing loss, limited language fluency or anxiety around unfamiliar technology. It also leaves out people who are tired, ill, grieving, rushed, carrying a child, trying to manage a disrupted journey or handling an urgent purchase.

Universal design does not mean every person must use the same tool. It means essential services should work for the widest possible range of people without requiring special pleading. A card with a clear tactile feature, a PIN process that does not depend solely on a touch screen, a staffed point of sale, a basic account with readable statements, and the option to use cash all contribute to that goal. A good system offers alternatives before someone has to disclose a disability or ask for an exception in public.

Cash has obvious accessibility limits. Counting change can be difficult for people with visual or cognitive impairments. Notes may be hard to handle for someone with limited dexterity. Counterfeit checks may create anxiety. These are reasons to improve note design, staff training and accessible digital tools, not reasons to assume that an app is universally easier. A facial-recognition login may fail for a person with a facial difference. A voice interface may be unsuitable for someone with a speech impairment. A smartphone-only flow can exclude a person whose device is shared, broken or inaccessible.

The most troubling design failures appear when a digital system removes the human route entirely. An automated payment terminal may tell a customer that a transaction failed but not explain what to do next. A chatbot may repeat a generic instruction without recognising that the user cannot complete it. A risk engine may require an additional biometric step without a workable alternative. These are not merely customer-experience defects. They turn money into a service that some people can access only through another person’s help.

AI can improve accessibility when it is used with care. Speech-to-text, translation, screen-reading support and simpler language can make financial services easier to use. AI can also create new barriers if it becomes the only support layer or if it misreads a person’s speech, intent or identity. Human assistance must remain available for people who need it, and the service should not penalise them for choosing it.

The test of an inclusive payment system is not whether it works smoothly for a confident smartphone user. It is whether a person can still pay independently when their body, device, language, circumstances or confidence do not match the default. Cash is one of the ways that independence remains possible.

Retail continuity depends on small practical decisions

Payment resilience is sometimes described through national systems, cyber standards and central-bank policies. It is also built through mundane decisions in shops. Does the business keep enough change? Can staff recognise when a terminal has lost connection? Is there a printed procedure for taking a payment during a disruption? Does the owner know how to communicate a temporary payment limit without creating panic? Can the business take a physical card offline where that is permitted? Does it still have a functioning cash drawer?

A merchant who says “cards only” may be making a reasonable choice on a normal day. The decision can become costly during a local failure. Customers who cannot pay leave without goods. Staff have to explain a problem they cannot fix. A queue forms while each person tries a different card, phone, app or bank. The business may lose a day’s turnover because an external provider is unavailable. A cash option can turn a total stop into a limited inconvenience.

Preparation does not require a small retailer to become a bank. It requires honest planning. Essential businesses such as grocers, pharmacies, fuel stations and transport operators should know which payments they can accept in a short outage and what conditions apply. They should maintain enough cash-handling capacity to use that option. They should train staff to follow a simple sequence: identify the disruption, state the options, protect customers from confusion, record what is necessary and restore normal controls afterward.

The Swedish central bank’s work on offline card payments for essential goods illustrates the value of concrete planning. It has focused on physical cards and PINs, not a vague promise that “digital payments will keep working.” The aim is to preserve a defined ability to buy essentials during disruption. Cash belongs in the same practical category. It is not a symbolic backup. It is a method that staff and customers can use immediately if it is actually accepted and change is available.

A practical resilience checklist for everyday payments

A checklist cannot eliminate every failure. It does make failure less chaotic. The key is to describe contingencies in practical terms rather than relying on slogans about “seamless” commerce. People judge payment resilience at the moment they need to buy something, not by the sophistication of the systems that failed behind the scenes.

AI will change people’s mental model of money

For centuries, the act of paying was visible. Someone counted coins, handed over notes, signed a receipt or wrote a cheque. Digital payments moved that act into a screen. Artificial intelligence may move it one step further, into a recommendation or instruction delivered by a conversational assistant. The consumer may not select a payment method in the ordinary sense. They may simply approve an agent’s proposal or set a rule and let the agent act.

That shift changes the mental model of money. If an assistant says it has found the cheapest option, the user may focus on the recommendation rather than the payment authority being granted. If a shopping agent reorders supplies automatically, the user may notice only after the goods arrive. If a finance assistant categorises spending and suggests a transfer, the line between advice and action can become blurred. A useful tool can quietly reduce the moments in which people actively consider what their money is doing.

There are benefits in delegation. A person with limited time may welcome automatic bill payments. Someone with a disability may find voice-based assistance liberating. A small business may reduce clerical work through automated invoice matching. The risk comes when delegation expands faster than understanding. A user needs to know whether the system can make a purchase, initiate a transfer, accept a subscription renewal, share transaction data or change a spending limit. Those permissions should be specific, visible and easy to withdraw.

AI systems also create a new form of authority bias. People may assume that a confident assistant has checked facts, compared terms fairly and applied the user’s interests. The assistant may instead reflect the incentives of a platform, rely on incomplete information, misunderstand a request or be manipulated by content on a merchant’s site. The more natural the conversation becomes, the easier it is to forget that a model is making probabilistic judgments rather than acting as a fiduciary.

Cash provides an important contrast. It does not make decisions. It cannot generate a recommendation, bundle an offer, accept a hidden subscription or infer a preference from past behaviour. A person has to decide to take it out and hand it over. That apparent limitation is a form of clarity. As AI makes payments less visible, cash preserves a direct relationship between intention and action.

Public policy should encourage tools that make delegation legible. Payment agents need clear limits, confirmation for unusual or high-value actions, straightforward records, a way to inspect decisions and a fast means of revoking authority. Providers should not use design tricks that turn a broad preference into an open-ended payment mandate. The future of payments may include agents, but people must remain able to understand what has been authorised in their name.

Records are useful until they become compulsory memory

A digital payment record often helps the person who made it. It proves that a bill was paid. It helps reconcile a business expense. It makes a refund easier to trace. It can support a complaint when goods are not delivered. It provides a history that can be valuable for budgeting, tax reporting and fraud detection. A serious argument for cash does not deny these benefits.

The problem begins when record creation is treated as a condition of ordinary life. A person may want evidence for a large purchase and discretion for a small one. They may want a searchable account history while also wanting to buy a gift, medicine or a newspaper without adding another item to a permanent commercial profile. These are normal preferences, not evidence of wrongdoing.

AI makes permanent records more consequential. A transaction history that once sat in a statement can now be classified, compared and combined with other data. Software may infer habits, preferences or vulnerabilities from a pattern that the customer never intended to disclose. The inference can shape advertising, credit assessment, fraud scoring or service eligibility. Even where regulations restrict misuse, the incentives to seek more data and derive more insight are strong.

The answer is not to reject records. It is to preserve choice and set strict boundaries around reuse. Payment providers should explain which data are needed to execute a transaction, which are retained for legal duties, which are used for fraud prevention and which are used for commercial purposes. Consent should not be buried inside a long contract. A person should be able to use ordinary payments without accepting unrelated profiling where that profiling is not necessary.

Cash keeps a space for transactions with no built-in commercial memory. It cannot produce an automatic proof of purchase, and a person who needs a receipt should ask for one. It does not solve every privacy question. It does limit the routine creation of data that can later be interpreted in ways the payer did not expect.

The design challenge for digital public money is particularly revealing. The ECB says that an offline digital euro is being designed to offer cash-like privacy, with offline payment details known only to payer and payee. That goal recognises that privacy is not a luxury feature. It is part of the public character of money. Cash remains the clearest benchmark because it starts with minimal routine data collection rather than trying to reconstruct privacy after the data system has been built.

Consumer protection has to match the speed of fraud

Digital payment systems can provide protections that cash cannot. Cardholders may dispute an unauthorised transaction. A direct debit can be challenged under defined rules. A bank can sometimes pause a suspicious transfer or trace funds through an account chain. Merchant records can support a refund claim. These mechanisms are valuable and should be strengthened.

Their limits become more visible as payments become faster. A victim of an impersonation scam may authorise a transfer personally after receiving a message that appears to come from a bank or relative. The payment can be technically authorised even though the decision was obtained through deception. If the money moves instantly through several accounts, recovery becomes difficult. The victim may be left to prove that they were manipulated while institutions debate where responsibility lies.

AI makes that dispute harder. A deepfake voice can create a believable emergency. A generated message can use a person’s name, employer and recent circumstances. A fraudster can test different stories at scale and learn which ones create urgency. Institutions should not respond by assuming every customer who makes an unusual transfer is careless. They should build friction into high-risk situations: warnings that are specific rather than generic, pauses where the risk is acute, easy reporting paths and rapid action once a scam is suspected.

The BIS and the CPMI have urged further use of advanced technologies, including AI, in fraud detection for cross-border payments. That work is necessary. It should be paired with clear consumer standards. A fraud model may identify a risk, but the customer needs to know what action is required, why the payment was stopped and how to reach a human who can resolve a mistake. A warning that merely says “be careful of scams” is weak when a person is being actively coached by a criminal.

Cash has a different consumer-protection profile. It is hard to reverse once handed over. It has no chargeback mechanism. It should not be portrayed as safer for remote transactions or as a remedy for every scam. Its value lies in reducing exposure to certain digital fraud routes, particularly those involving fake links, account takeover, card details and pressured transfer authorisation.

A resilient consumer-protection system needs both strong digital remedies and a form of money that does not expose every small face-to-face purchase to remote fraud techniques. The goal is not to choose between protection and cash. It is to ensure that neither is lost.

Identity systems should not become payment gates

Digital identity is becoming central to online banking, public services and commerce. Properly designed, it can reduce repetitive paperwork and make it easier to prove who someone is without sharing more personal data than necessary. It can also support fraud prevention by making account opening and recovery more secure.

The danger comes when identity verification becomes a universal gate for transactions that do not need it. Buying a newspaper, a sandwich or a local bus ticket should not normally require a person to prove a formal digital identity. Nor should an individual who has lost a phone or is waiting for documents be treated as unable to use money at all.

AI can make identity systems more capable and more intrusive at the same time. Automated document analysis, liveness checks and behavioural signals can speed up verification. They can also create opaque failures. A customer may be told that a document “could not be verified” without a useful reason, or that a risk score is too high without a clear route to correct a bad record. These systems may be particularly difficult for people whose names, addresses, faces or documents do not fit the patterns on which the model was tested.

Cash offers an ordinary transaction path that does not require a digital identity event. The merchant may need to comply with specific laws in particular situations, but a small cash purchase does not normally involve account opening, biometric matching or profile scoring. That is one reason physical money remains important in a society where identity checks are expanding.

The policy principle should be proportionality. Verify identity when the activity genuinely requires it: opening an account, accessing sensitive records, borrowing money, completing a high-risk transfer or meeting a legal obligation. Do not turn routine payments into continuous identity audits merely because technology makes it possible. Digital identity should make essential services easier to reach, not create a new category of people who are technically present but unable to transact.

People need the ability to pay before every part of their digital identity is perfect. Cash protects that ability, while better identity systems should reduce the occasions on which a person needs to use it as a workaround.

The cash network needs clear public responsibility

Cash access is often left to a patchwork of commercial choices. A bank closes a branch. Another bank reduces cash services. An ATM operator relocates a machine. A retailer stops accepting notes at some tills. No single decision appears decisive. The cumulative result can be a town where people must travel far to withdraw cash, a small business where deposits are costly, or a neighbourhood where the only cash machine charges a fee.

That outcome is not inevitable. It reflects a policy choice to treat cash distribution solely as a private service rather than a component of public infrastructure. A country does not need every bank to provide every service in every location. It does need a clear answer to a basic question: who is responsible when no commercial provider maintains reasonable access?

Possible models include shared ATM networks, universal-service obligations, cash-back arrangements through shops, public subsidies for rural access, minimum service standards and obligations on major banks to coordinate. Each carries trade-offs. Cash-back through a retailer can be useful but depends on the shop’s own cash stock. A shared ATM network can reduce duplication but may still leave remote areas under-served. A public obligation can protect access but must be defined carefully so it is workable.

The ECB’s cash strategy treats access to and acceptance of cash as necessary conditions for preserving the freedom to choose how to pay. The practical consequence is that policy cannot stop at statements of principle. It needs measurable standards: distance or travel time to cash withdrawal, the availability of deposit services for small businesses, uptime expectations, transparency around fees and a process for communities to raise concerns.

Public responsibility also includes education. People should know which notes are legal tender, how to identify a suspect counterfeit, where to report a damaged note and what to do if a merchant refuses cash. Businesses need clarity on their obligations and permitted exceptions. Confusion helps neither side.

Cash will remain a real choice only if its infrastructure is managed as a public-interest system rather than allowed to disappear through a series of uncoordinated local decisions.

Payment resilience should be tested, not assumed

A digital payment system can appear strong for years and still fail under conditions that have never been tested. A bank may have backup data centres but not enough customer-support capacity when thousands of users receive a fraud alert at once. A retailer may have an offline-card procedure but staff may not know where the physical cards, manuals or limits are. A municipality may have an emergency payment plan but no way to communicate it when its website and app are unavailable.

Testing should include realistic scenarios. A local telecoms failure. A payment-provider outage on a busy weekend. A ransomware incident that forces a bank to restrict some services. A large storm that disrupts power and transport. A false fraud-model update that blocks a class of ordinary transactions. A coordinated phishing campaign that causes customers to call in at the same time. The goal is not to predict every event. It is to discover whether the fallback works when people are stressed.

Tests should involve the whole chain. Banks cannot assess resilience alone if merchants cannot accept the available alternatives. Telecoms companies cannot assess it alone if payment terminals have no backup. Local authorities cannot assess it alone if residents have no access to cash. The Riksbank’s recent payment-preparedness work is notable because it joins central-bank planning with market actors and focuses on essential goods and actual methods of payment.

Cash should be included in these exercises. Are local cash machines stocked? Do shops have change? Can a branch or cash centre operate under the scenario? Do customers know that cash will be accepted? Are there sensible limits to prevent panic withdrawals while still allowing people to obtain what they need? A country that has not tested these questions does not know whether cash is truly available in a crisis.

The same standard applies to AI. A fraud system should be tested for false positives across different customer groups. An automated support system should be tested for escalation when the customer cannot resolve a problem alone. An AI payment agent should be tested for manipulation, unclear authority and accidental overspending. Model performance on a laboratory dataset is not a substitute for resilience in the real world.

Trust grows when institutions can show that their backups work under pressure, including the oldest backup of all: the ability to use cash.

A policy framework for payment pluralism

The case for cash is strongest when it is framed as part of a wider payment policy rather than as an isolated demand. A country needs more than notes and coins. It needs secure digital systems, prompt fraud response, accessible bank accounts, clear consumer rights, reliable telecoms, physical cards, human support and sensible rules for new forms of money. The aim is plurality: several usable routes to payment that do not all fail for the same reason.

That framework should begin with a public commitment that cash remains available and accepted for ordinary face-to-face transactions. Exceptions may be justified for remote commerce, unattended equipment, security-sensitive locations or clearly agreed special arrangements. The default should not be a quiet shift to cash refusal simply because a merchant finds digital operations easier.

Second, governments and regulators should set access standards. The standard should cover withdrawals and deposits, not only ATMs. It should account for rural areas, low-income communities, older residents, people with disabilities and small businesses. It should include fee transparency and a mechanism for intervention where local access falls below a reasonable threshold.

Third, digital payment resilience should be treated as a shared national responsibility. Essential merchants should be able to accept at least one payment method during a meaningful disruption. Physical cards and PINs deserve more attention because they do not rely on a charged smartphone. Cash stocks and change should form part of local emergency planning. Offline systems should be designed with honest limits rather than marketed as universal solutions.

Fourth, AI governance in payments should focus on accountability. Automated fraud tools should have clear human-review routes. Customers should be told when a decision was automated and should be able to challenge a harmful error. Payment agents should operate under specific, revocable authority. Institutions should monitor model bias, data quality, third-party dependence and concentration risk. The BIS has warned that AI can raise complexity and make risk harder to observe, especially where systems rely on opaque models, unstructured data and third parties.

Fifth, privacy needs practical protection. Data minimisation, purpose limitation, transparent permissions and restrictions on transaction-data reuse should apply to private systems. Public digital money should be built around privacy by design. Cash should remain available as the low-data option for lawful daily transactions.

Finally, public communication needs to be calmer and more honest. Cash is not a cure for fraud, and digital payments are not a plot against freedom. Both claims are simplistic. People need to understand the genuine trade-offs: speed against reversibility, convenience against dependency, data records against privacy, automation against human control, and efficiency against resilience.

Payment pluralism is not indecision. It is deliberate system design for a world in which technology brings both capability and fragility.

The future should leave room for refusal

The most valuable freedom in a payment system is not the ability to choose from a long menu of apps. It is the ability to decline a particular form of mediation. A person should be able to say: I do not want to use my phone for this. I do not want to create another data point. I do not want a shopping assistant to act on my behalf. I do not want to wait for a fraud model to decide whether I can buy a meal. I do not want to depend on a platform that has failed today.

Cash makes those refusals practical in face-to-face life. It does not remove every dependency. People still depend on shops, employers, transport, the state and the broader monetary system. It does remove the need to depend on a specific device, account, network or automated assessment at the moment of an ordinary exchange.

That freedom has value even for people who rarely use it. A person who prefers cards may still want cash during travel, a local outage or a period of account trouble. A merchant who prefers contactless payments may still want notes when terminals fail. A parent who teaches a child to budget may still use an app for school payments. A digitally confident citizen may still dislike the idea that every purchase is permanently available for analysis.

AI will make payment systems more capable. It may make them safer, faster and more personalised. It will also make them harder to inspect from the outside. Decisions that once appeared as a human judgment may be embedded in models, risk scores, hidden thresholds and automated workflows. The appropriate response is not fear of technology. It is to preserve a zone of everyday economic life where technology is optional.

Cash is that zone. It is public money in a form that a person can understand by looking at it and use by handing it over. It has faults, costs and limits. It also has a quality that will become rarer as automated systems expand: it works without asking a machine to interpret the person using it.

The meaning of cash in the age of AI is therefore not resistance to the future. It is the preservation of a human fallback inside it.

Its limitations are visible too: it can be lost, stolen, miscounted or refused in a situation where a merchant has not prepared to accept it. Nothing about cash removes the need for good banking, good regulation or good digital services. Its value lies in keeping one basic act—paying another person face to face—outside a growing web of software, identifiers and remote approval.

That is why the debate should not be framed as a contest between an old habit and a new technology. It is a question about institutional design. A payment system that works only for people with a current smartphone, a live connection, a recognised device, a functioning account and a favourable risk score is not fully public. It has converted participation in everyday commerce into a conditional service. Cash is the residual form of payment that does not ask citizens to qualify in those ways before they can buy an ordinary good.

The public often notices the importance of a fallback only after losing it. A town may not feel the loss of a bank branch until the nearest cash withdrawal requires a long journey. A shopper may not think about cash acceptance until a terminal is down. A person may not care about account-recovery procedures until a stolen phone leaves them unable to pay for transport. By then, the debate has moved from preference to dependency. The practical importance of cash becomes clearest when the digital option is unavailable at the moment of need.

A strong digital payment system should make that moment rare. Banks and providers should invest in redundancy, secure systems, accessible support and rapid error correction. Public authorities should set resilience standards and enforce them. But even very good systems fail sometimes, and AI adds new types of failure that are not cured by better broadband alone. An automated fraud control can stop a legitimate purchase. An identity check can reject a genuine user. A customer-service bot can misunderstand a crisis. A cloud-service interruption can affect many firms at once. Cash exists outside those specific failure modes.

There is a tendency to call every retained analogue option inefficient. That description confuses low frequency with low importance. Emergency exits are inefficient in ordinary use. Paper maps are inefficient until a device loses power. Printed contact numbers can look redundant until a network is inaccessible. Cash plays a similar role in payments. It may account for a smaller share of transactions than in the past while carrying a large share of fallback value. The fact that people use it less during ordinary conditions does not prove that society can safely lose it.

The relevant measure is therefore not only transaction share. It is substitutability. When a card fails, what can the customer use instead? When a wallet provider suspends an account, can the person still buy food? When an AI-driven fraud review delays a transfer, can the household meet a small urgent expense? When mobile service drops during travel, does the passenger have a ticket option that still works? A payment system with good answers to these questions has more resilience than one that merely records high digital adoption.

Cash also gives people a way to separate economic life from technical competence. A person may be intelligent, responsible and financially cautious yet still have difficulty with a complex app or a biometric prompt. They may not speak the language used in a support flow. They may have a disability that makes a touch-screen process hard. They may be in a place where service is poor. The ability to use money should not depend on passing an unannounced usability test every time a person makes a purchase.

The right to choose cash does not mean a right to force a specific merchant to take an unlimited quantity of notes in every circumstance. Reasonable exceptions exist. Remote sellers cannot receive physical currency through the internet. A machine without staff cannot handle change. Security restrictions may apply in sensitive locations. A shop may agree a different method with a customer in advance. These practical limits do not weaken the general principle. In ordinary face-to-face trade, a public currency should normally remain usable unless there is a clear reason otherwise.

This distinction matters because cash refusal can spread through design rather than direct prohibition. A restaurant replaces its staffed till with table ordering. A transport operator moves tickets into a mobile application. A parking system takes payment only through a QR code. A public counter directs every visitor to an online portal. A retailer opens self-checkout lanes without cash capacity. Each decision can be presented as an operational improvement. The result may be a daily economy in which people are free to use cash only in the remaining places that have not redesigned them out.

AI can accelerate that drift. A business may use automated demand forecasts, labour scheduling and customer analytics to reduce physical staffing. A vending system may recognise customers through a phone account. A conversational shopping tool may direct a person to a payment option linked to a digital profile. None of these practices is automatically wrong. The issue is cumulative. If all physical transactions are rebuilt around authenticated digital access, cash acceptance becomes harder not because anyone voted against cash, but because there is nowhere left to hand it over.

That is a policy problem precisely because markets do not price every public consequence. A merchant can calculate terminal fees and staff time. It may not calculate the cost imposed on an older customer, a visitor, a person leaving an abusive household, a child without a phone or a worker whose account has been temporarily blocked. Those costs fall on individuals and communities. Regulation has a role when private efficiency produces public exclusion.

The European policy debate has recognised this. The European Commission’s proposal on the legal tender of euro cash was intended to safeguard acceptance and access while permitting limited exceptions. The proposal reflects an understanding that legal tender cannot be meaningful if cash is regularly excluded in practice. The Eurosystem’s cash strategy likewise treats availability, accessibility and acceptance as linked conditions. A currency is not truly public when citizens must search for the few places willing to accept it.

This does not require a single rigid rule for every economic activity. It requires a presumption that ordinary commerce should not force people into a particular technology stack. A country can support new payment forms while setting standards for cash access. It can make digital services easier while retaining staffed routes. It can allow automated checkouts while ensuring that a cash-capable option exists nearby. The aim is to prevent the most common payment environments from becoming private membership systems by default.

Cash is also a form of practical interoperability. A note does not require the payer and merchant to use the same bank, phone operating system, wallet app or payment network. It does not matter whether the buyer’s device is old, foreign, uncharged or incompatible. The buyer and seller must agree only on the currency and the amount. In an economy full of proprietary interfaces, that simplicity is unusually powerful.

Digital payment interoperability has improved, but it remains partial. A card may work across borders yet rely on a scheme accepted by the merchant. A wallet may work only on certain devices. An instant-payment application may require the recipient to use a compatible bank. A digital identity credential may not travel easily across systems. These constraints often become visible at the edge: a small merchant, a cross-border worker, an offline area, a person with an older phone, or a customer who has fallen outside a provider’s risk tolerance.

Artificial intelligence may increase rather than reduce this fragmentation. Companies will offer proprietary assistants, data ecosystems and automated shopping flows. Each will work best inside its own network. An agent trained on one platform may be less capable of comparing alternatives outside it. A payment recommendation may be influenced by commercial arrangements that the user does not see. The danger is not only that people become locked into a provider; it is that they stop noticing that lock-in because the assistant makes the experience feel smooth.

Cash is a blunt antidote to that problem. It does not solve online or remote transactions, and it cannot offer every digital service. But in the physical world it remains neutral infrastructure. It treats all lawful holders of euro notes alike. A note does not ask which platform the customer belongs to.

There is a related issue of bargaining power. A person who can pay only through one account or one app has little leverage when that provider changes its terms. The provider may introduce a fee, collect more data, require an update, alter its support process or withdraw a feature. The customer can complain, but leaving may be difficult when the service is needed for daily life. A person who still has access to cash retains at least a limited ability to transact outside that relationship.

Businesses face the same problem. A small shop that relies entirely on card acceptance may be exposed to changes in merchant fees, dispute rules or terminal contracts. Large retailers can negotiate. A family business in a small town often cannot. Cash does not remove tax, accounting or security duties, but it gives the business a channel that is not governed by a private payment provider’s real-time decision. That channel can matter during a contract dispute, outage or sudden change in service terms.

This does not mean cash is free of power. Banks determine where cash can be withdrawn and deposited. Central banks set the currency framework. Governments regulate legal tender and crime. Merchants decide whether they can handle notes safely. The difference is that the structure is more public and more legible. The rules of a banknote are not hidden inside a recommendation system or an account-risk model. A person can see the instrument and understand the transaction without accepting a complex digital agreement.

The clarity of cash has value in disputes. A digital payment can produce a detailed record, which is helpful when goods are not delivered or a charge is contested. It can also produce a complex chain of responsibility. Was the payment authorised? Was the device compromised? Did the customer ignore a warning? Did the merchant submit the transaction correctly? Did a model classify it as suspicious? Did a third-party service fail? The more intermediaries involved, the more difficult it can be for a customer to understand which party controls the remedy.

Cash simplifies the moment of exchange. The buyer pays and the seller accepts. A dispute about the goods may remain, but the payment route is not hidden behind multiple institutions. This is one reason why cash is not merely a nostalgic object. It is a simple protocol. Simplicity reduces some risks even as it removes some protections. A mature payment policy should respect both sides of that trade-off.

The issue becomes sharper as software agents are given permission to spend. An agent may be useful when it reminds someone of a bill or finds a lower price. It becomes more consequential when it is allowed to place an order, renew a service or move money. The user needs to know whether the agent is providing information or exercising authority. The distinction can be obscured by conversational interfaces that make a system sound like a helpful colleague rather than a service with limited, probabilistic capabilities.

A well-designed payment agent should have narrow, visible permissions. A person might allow it to pay one named bill up to a specified amount, but not to change bank details. They might allow it to make a recurring purchase from a named merchant, but not to accept a higher price or a substitute product. They should be able to inspect every action, revoke authority immediately and receive a meaningful alert when the agent encounters an unusual situation. These safeguards are not a technical luxury. They are the minimum conditions for informed consent.

The same logic applies to merchant-side agents. A seller may use AI to approve refunds, detect suspicious orders or adjust fraud controls. Those systems should not turn every customer complaint into an automated dead end. A person who believes a payment was wrongly rejected or a refund wrongly withheld needs an accessible human route. The provider should be able to explain the decision in plain language and correct it quickly where the evidence warrants.

Automation should not turn the exercise of financial rights into a puzzle. The more decisions are made at machine speed, the more important it becomes to preserve a slow, accountable path for correction. Cash provides one form of escape from some digital decisions, while human review provides another. Both belong in a payment system that takes people seriously.

The language used around AI often makes it sound more certain than it is. A fraud model may assign a probability, not establish a fact. It may detect correlations that have no clear explanation. It may perform well on transactions similar to those in its training data and poorly on situations that are new, unusual or poorly represented. Payment providers know this in principle, yet customers often encounter the output as an absolute instruction: declined, blocked, verified, restricted.

Institutions should resist treating model outputs as unquestionable. They should test whether particular groups are more likely to experience false positives, whether changes in customer behaviour are being mistaken for fraud, and whether model performance worsens during an outage, crisis or market shift. They should audit third-party tools rather than assuming that a vendor’s claim of accuracy answers the governance question. They should also document how a customer can challenge a decision without having to understand the model itself.

Such safeguards are especially important where payment access affects urgent needs. A model that pauses a large, unusual transfer may protect a victim from a scam. A model that blocks a customer’s only way to pay for travel, medicine or accommodation may create a different kind of harm. The response should be proportional. A bank can flag and confirm a transaction without leaving a person stranded. It can set temporary limits while preserving access to a small emergency amount. It can offer a human decision-maker rather than a generic chatbot script.

Cash is not a technical solution to model governance. It does, however, limit the damage of a narrow class of false positives. A person who has enough cash for immediate essentials can tolerate a digital review more safely than one whose entire ability to transact has been suspended. The existence of cash does not excuse poor bank processes. It gives the individual more room while those processes are corrected.

The need for room matters in a society where people are expected to respond instantly. Fraudsters create urgency. Platforms reward quick decisions. Retail systems turn hesitation into a lost sale. Social media can turn a rumour into a payment request within minutes. Cash introduces a small pause. A person has to decide whether to withdraw it, carry it and hand it over. That friction is not always convenient, but it can protect judgment.

Behavioural economics is often used by businesses to reduce friction and increase conversion. One-click payments, saved cards, automatic renewals and personalised offers all reduce the distance between impulse and purchase. AI can tailor that reduction to an individual: the time of day, the wording, the item, the discount, the apparent scarcity and the recommended payment plan. Consumers gain convenience, but they also face a commercial environment that can learn which prompts are most likely to work.

A cash budget offers a simple countermeasure. A person can decide that a fixed amount is available for discretionary spending and leave cards or phone payments for planned essentials. This is not the right technique for everyone. It can be inconvenient, and people with mobility or safety concerns may prefer digital tools. It remains useful because the boundary is physical. A recommendation engine cannot expand it. A subscription cannot draw from it overnight. A merchant cannot turn it into a recurring authorisation.

The same principle can support households trying to teach children about money. A child who handles a small amount of cash experiences trade-offs directly. They learn that saving requires not spending, that change matters, and that a price is not just a number on a screen. Digital financial education remains necessary, especially as children encounter online games, subscriptions and in-app purchases. Physical money gives that education a concrete starting point.

Cash can also help people whose spending is affected by anxiety or compulsive behaviour. It is not therapy and should never be offered as a cure. A person managing a difficult relationship with shopping, gambling or debt may benefit from professional support, self-exclusion tools and account controls. Still, a deliberate cash limit can provide a boundary that is easier to understand than a screen full of available credit or stored payment credentials. Choice of payment method can be part of financial self-management, not just a question of convenience.

Privacy deserves the same practical treatment. The claim is not that cash makes a person invisible. Shops may use cameras. Law enforcement can investigate crimes. Large cash transactions can trigger rules. A person may need a receipt. The point is narrower: ordinary cash use does not automatically generate a detailed transaction record held by a chain of private and public actors.

For many people, that difference is ordinary rather than ideological. They may want to buy a gift without leaving it in a shared account history. They may want to pay a neighbour back without turning a personal exchange into a platform event. They may prefer not to add a small purchase to an advertising profile. They may simply dislike the idea that every daily movement can be reconstructed later from payment data. Those preferences do not require suspicion of banks or rejection of digital commerce.

AI makes the record more revealing because it can interpret patterns that a human observer would not notice. A long transaction history can suggest routines, relationships, employment changes, travel, health-related spending, household composition or financial stress. Some inferences will be wrong. Wrong inferences can still affect a person if they influence marketing, credit offers, fraud scores or customer treatment. The risk is not only that data are stolen. It is that data are used exactly as designed, but for purposes the customer never understood or agreed to.

Data-protection law provides important safeguards, including principles of minimisation, purpose limitation and proportionality. Those principles need active enforcement in payment markets. It should not be enough for a company to say that it “uses data to improve services.” Customers deserve a clear account of what information is needed to process a payment, what is retained for legal obligations, what is used to prevent fraud and what is used for commercial profiling. A person should not have to read a long policy to know whether buying a coffee is also creating a marketing signal.

Cash preserves a low-data alternative. It does not require everyone to reject digital records. It allows a person to decide that some transactions do not need to become part of a permanent machine-readable history. That choice becomes more rather than less relevant as AI increases the value of transactional data.

The argument has a public-finance dimension as well. Governments need tax revenues and must combat serious crime. Digital records can assist legitimate investigations and compliance. The existence of a cash option does not prevent targeted enforcement. It does prevent a situation in which every citizen’s ordinary spending becomes visible by default on the theory that somebody might misuse money. Democratic systems normally distinguish between broad surveillance and evidence-based investigation. Payments should be treated with the same care.

A proportional approach would focus scrutiny where risk is genuinely high: suspicious patterns, large transactions, professional intermediaries, credible intelligence and legal thresholds. It would not treat ordinary cash use as a sign of bad faith. Nor should it assume that digital flows are automatically clean. Online fraud, mule-account networks, ransomware payments and false identities all show that traceability is useful but not self-executing. Institutions need capacity, cooperation and lawful authority to turn records into action.

The public can reasonably expect two things at once: serious work against financial crime and a private space for lawful everyday exchange. These goals are not incompatible. They require careful rules and a refusal to turn technological capability into automatic permission. Cash is part of that refusal.

The environmental argument also benefits from restraint. Notes and coins have material costs. They are manufactured, transported, stored, secured and replaced. Digital payments use devices and networks that must be built, powered and renewed. AI systems use computing resources that may be substantial, particularly when models are trained or used at scale. A credible comparison needs to look at lifecycle impacts rather than presenting one form of payment as weightless because it appears on a screen.

There is no single environmental score for “cash” or “digital.” A note used hundreds of times has a different footprint from one quickly removed from circulation. A payment card used for years differs from a frequently replaced smartphone. A local cash purchase differs from a data-intensive online purchase that triggers delivery, returns and advertising activity. The right policy response is to make every payment system less wasteful: durable notes, efficient cash distribution, long-lived devices, repairable hardware, lower-energy data infrastructure and transparent reporting.

Cash should not be preserved by ignoring its environmental costs. Digital payments should not be promoted by ignoring theirs. The public deserves honest comparisons that identify what is being counted. Resilience, inclusion and privacy cannot be reduced to a single carbon metric, but environmental impact should be assessed openly alongside them.

There is a cultural dimension to physical money that is often dismissed too quickly. Banknotes are a shared object. They are recognisable across income groups, regions and generations. A person without a bank account can hold the same currency as a person with a complex investment portfolio. That commonality matters because money is not only an accounting entry. It is a social promise that people can exchange goods and services on equal terms.

Digital payment systems can be equally universal in theory, yet in practice they often divide users by device quality, account type, data access, credit history and platform membership. A premium customer may receive instant support and extra features. A customer with a basic account may face stricter limits or less human help. A person who cannot satisfy a verification process may be excluded entirely. Cash does not erase economic inequality, but it does not amplify those distinctions at the point of a small payment.

The public nature of cash is particularly relevant during anxiety about financial institutions. When rumours spread online, people may struggle to distinguish a real bank notice from a fabricated one. A deepfake video can make a false claim feel credible. Automated systems can amplify a message before institutions have time to respond. In such conditions, trust in a basic, tangible form of public money can stabilise behaviour. It does not solve a banking crisis, but it gives people something they understand without needing to judge a viral claim about an app.

This is another reason central banks need to communicate clearly. They should explain the role of cash, the security features of notes, the availability of withdrawal facilities and the limits of rumours. Banks should have ready communication channels that do not depend entirely on the compromised or contested platform where misinformation is circulating. Public confidence is not created by reassurance alone; it is created by systems that work and information people can verify.

Cash contributes to that confidence because it is legible. A note has visible denominations and public security features. It does not require a customer to trust a code update or a recommendation algorithm. The physical object does not replace institutional trust, but it gives that trust a tangible form.

The international context strengthens the case for preserving several payment routes. Payment systems are increasingly linked across borders. That connection supports travel, trade, remittances and online commerce. It also introduces dependencies on foreign networks, cloud services, software vendors and technology standards. A country that has many digital payment options may still be exposed if those options rely on the same external infrastructure.

Europe has been examining these dependencies as it considers a digital euro and more integrated payment systems. The aim is not isolation. Cross-border payments are necessary in a connected economy. The aim is to ensure that a political or commercial disruption outside a country does not remove the ability of people inside it to conduct ordinary commerce. Cash is one component of that autonomy because domestic notes can circulate locally even when external services are affected.

The same logic applies to private technology companies. They may be innovative and reliable partners. They may also change strategy, leave a market, alter a product or face an incident that affects millions of users. No country should build its everyday payment capacity on the assumption that every global provider will always be available on terms that meet domestic needs. Public infrastructure needs more than one route.

This is not an argument for national isolation or for rejecting global payment networks. It is an argument for contingency. A state should know which capabilities it controls, which it contracts for, and which it can use if the usual digital layer is disrupted. Cash is a domestic public payment rail that remains usable when international or private digital rails are unavailable.

The resilience question becomes more serious when considering essential services. Food shops, pharmacies, fuel stations, public transport, hospitals and emergency shelters should not discover during a crisis that their payment arrangements have no fallback. Their plans should identify what can be sold, what records must be kept, how price controls and consumer rights will be maintained, and which payments work without ordinary connectivity. Cash needs to be part of those plans, alongside offline-capable cards and manual procedures where they are lawful and safe.

Preparedness should be proportionate. A small village shop does not need the same arrangements as a hospital or fuel terminal. It does need to know whether it can take cash and make change if a terminal fails. A city transport system may need more than one ticket route, including staffed or cash-capable options. A bank should have a plan for customers who cannot use an app because of a fraud hold or outage. A municipality should avoid relying on a single app for emergency assistance.

Exercises can reveal weaknesses before they become public failures. What happens when mobile networks are congested? What happens when people receive a large number of false fraud alerts? Can a merchant accept cash if its usual deposit service is unavailable? Can a customer find accurate information without logging into an app? Can a vulnerable person obtain a small amount of money quickly if their card has been blocked? The answers should be tested in realistic conditions, not assumed from a policy document.

AI can contribute to those exercises by simulating demand, identifying bottlenecks and analysing anomalies. It should also be treated as a dependency. What if the fraud model is unavailable? What if an AI support system gives bad advice during an outage? What if a model update incorrectly blocks a class of payments? A resilient organisation plans for failure in the very tools it uses to manage failure.

The household role in preparedness should be described without alarm. Keeping a modest cash reserve in secure storage is a practical step, not a call to panic or hoard. The amount will vary by household needs, location and ability to access other support. Mixed denominations are often more useful than a single large note. People should also know where their essential documents are, how to contact their bank through verified channels and how to use a physical card if a phone is not available.

Cash is one element of a broader personal plan. A person may keep emergency contacts on paper, carry a charged battery pack, know the route to a cash machine and review account-security settings. The goal is not self-reliance in every crisis. It is to reduce avoidable dependence during the first hours of a disruption. Public institutions still have responsibilities; household preparation cannot substitute for them.

The strongest argument for cash is therefore neither fear nor nostalgia. It is a disciplined recognition of trade-offs. Digital systems offer reach, speed, records and consumer protections that physical cash cannot match. Cash offers direct settlement, everyday privacy, interoperability, low technical dependence and a fallback when automated systems fail. These features are complementary. A society that keeps both is not trapped between past and future. It is better equipped for both.

The phrase “cashless society” can sound like a neutral description of consumer choice. It becomes less neutral when people who prefer or need cash are unable to obtain it or use it. A choice made by an affluent, connected consumer is not the same as a choice imposed on someone who has no device, no account, no reliable signal or no confidence in a fraud-prone digital environment. Public policy needs to distinguish voluntary adoption from forced dependence.

A cashless economy may still contain cash in a technical sense, but if it cannot be withdrawn conveniently, accepted routinely or deposited safely, it has ceased to function as a practical option. That is the condition policymakers should avoid. The task is not to preserve cash’s old share of payments. It is to preserve its usefulness.

The same standard should be applied to future public money. A digital euro, if it is issued, should expand the ability of people to use central-bank money in a digital environment. It should not become a pretext for reducing physical cash access. It should offer strong privacy protections, accessible forms of use and genuine offline capability where technically possible. It should be accompanied by clear safeguards against unnecessary programmability and surveillance. Most importantly, it should operate alongside notes and coins rather than setting the stage for their withdrawal.

The ECB and European Commission have repeatedly stated that the digital euro would complement cash, and the legal-tender proposal aims to protect access to and acceptance of euro notes and coins. Those commitments deserve to be judged by implementation. People will look at the nearest cash machine, the shop counter, the public ticket machine and the bank’s response to an account problem. Principles become credible only when they are visible in daily life.

For Slovakia and other euro-area countries, this is a practical test. Are cash withdrawal and deposit services available across urban and rural areas? Do essential merchants have payment-continuity plans? Can an older person or a visitor pay for transport without a smartphone? Does a person with a blocked digital account have a fast route to support and a way to buy essentials in the meantime? Are cash refusals monitored rather than assumed to be harmless? These questions matter more than slogans about innovation.

A payment system worthy of public trust should allow people to choose the method that fits the moment. A phone can be convenient when it is charged and connected. A card can be useful when the customer wants a record and a dispute process. An instant transfer can be appropriate for sending money to a known recipient. Cash can be the sensible choice when privacy, independence or offline use matters. The public should not have to defend that choice as a refusal of progress.

The real risk is not that cash survives. The real risk is that society loses it by neglect—one branch closure, one cashless till, one app-only service, one untested contingency plan at a time—until the option disappears before people realise what it was doing for them. By then, restoring it is more expensive and harder than preserving it would have been.

That outcome can be avoided. Governments can set access and acceptance standards. Central banks can maintain a sustainable cash cycle. Banks can cooperate on withdrawal and deposit infrastructure. Merchants can retain practical cash capability. Digital providers can build offline options, protect privacy and offer human support. Regulators can insist that AI systems remain contestable, explainable enough for meaningful review and subject to operational-resilience controls. Consumers can keep a small amount of cash and learn the differences between payment methods.

None of these steps requires rejecting AI. They make AI’s use in payments more accountable. A fraud system is easier to accept when a person can reach a human and still meet immediate needs. A digital wallet is easier to trust when it does not monopolise access to money. A public digital currency is easier to support when physical public currency remains protected. Innovation is stronger when people retain a way to opt out of a particular channel without opting out of economic life.

Cash remains important in the age of AI because it is the payment method least dependent on AI’s assumptions. It works without asking a model whether the transaction fits a pattern, without turning a lawful small purchase into a data point, and without requiring a person to maintain a constant connection to private digital infrastructure. It is not a complete payment system and should not be treated as one. It is the part of the system that remains available when other parts become too complex, too intrusive or temporarily unavailable.

That is a modest role, but it is a fundamental one. Money is not merely a product feature. It is the medium through which people obtain food, travel, shelter, care, work and independence. A society that lets those basic acts depend on a single digital path accepts more risk than it needs to. Cash keeps another path open.

The first duty is to make payment choice measurable. Governments often make broad statements about inclusion, resilience and freedom to pay, but those promises mean little without service standards that citizens can observe. A standard might specify how far people should normally need to travel for free cash withdrawal, whether small businesses can deposit takings locally, how often cash machines may be unavailable, and which public-facing services must offer a non-smartphone payment route. It should include rural areas and neighbourhoods that private networks find less profitable. Publishing the results would turn a vague commitment to cash into something people can hold institutions to account for.

Merchants need equally clear expectations. A rule that describes cash as legal tender but permits routine cash refusal through unattended systems, app-only ordering or cashless self-checkouts will not preserve a practical choice. Physical businesses should be able to use exceptions where safety, distance or the nature of the service makes cash handling unreasonable. The starting assumption should remain that a customer standing at a staffed public counter can use euro notes and coins. This is not an operational burden imposed for symbolism. It keeps everyday commerce open to people whose digital access is interrupted or who do not wish to turn every purchase into a platform transaction.

The same discipline applies to artificial intelligence. Financial institutions should distinguish between a model that recommends a review and a model that effectively denies someone access to money. The first is a useful security tool. The second needs stronger safeguards: clear notice, a time limit for review, a human contact route and a way to preserve access to essential spending. A fraud system should not create the very crisis it is intended to prevent. Banks will not be able to disclose every technical detail of a model, but they can explain the practical reason for a hold and the steps required to resolve it.

Performance measures should reflect that responsibility. It is not enough to count how many fraudulent transactions were stopped. Providers should also examine false positives, the time customers spend locked out of funds, the number of cases that a chatbot fails to resolve, the quality of escalation for people with accessibility needs, and the rate at which customers recover losses after authorised-push-payment scams. These indicators show whether security is working for the public rather than merely reducing an institution’s losses. In an AI payment system, the quality of the remedy matters almost as much as the quality of the detection.

Data rules require similar attention. Payment providers have legitimate duties to detect fraud, meet anti-money-laundering obligations and resolve disputes. Those duties do not create an unlimited licence to reuse transaction histories for unrelated commercial profiling. A person should be able to tell which data are necessary to execute a payment, which data are retained by law, and which are optional. Permission for optional uses should be specific rather than obtained through a blanket agreement that few customers can reasonably understand. Cash is relevant here because it preserves an ordinary option for lawful low-value payments with no default commercial dossier attached.

The structure of consent will become more important as payment agents appear in everyday life. A customer should not have to choose between giving an assistant unrestricted spending authority and abandoning the service. Permissions can be narrow: one bill, one merchant, one category, one monthly ceiling, one period of time. The user should see what information the agent can access, whether it receives payment incentives from merchants, and whether it is allowed to change a supplier or accept a higher price. Reversing a permission should be as simple as granting it. A person who cannot understand the authority being delegated has not given meaningful consent.

Consumer protection must adapt before these tools become routine. Existing rules often ask whether a transaction was authorised. Agent-mediated payments may require a more precise question: authorised by whom, under what instructions, with what ability to detect manipulation, and under what provider responsibility? A customer should not carry the full loss merely because a model or assistant technically followed an ambiguous instruction. Providers that market agents as trusted helpers will need to accept duties that match that trust. Otherwise, the commercial benefit of automation will be paired with the consumer cost of every misunderstanding.

Public procurement can shape this market. When a city buys ticketing, parking, social-benefit or public-fee systems, it should ask whether residents can pay without a smartphone, whether the service has an outage mode, whether data are reused, and whether support remains available from a person. These requirements should appear in contracts, not merely in accessibility statements. Public money should not finance systems that make the public currency unusable or that exclude citizens who cannot clear a digital identity hurdle on demand.

Businesses also need practical incentives to retain cash capability. The cost of a cash register, insurance or a deposit arrangement is visible. The cost of losing trade during a card outage is visible only after it happens. Local business groups, banks and municipalities can reduce the burden through shared deposit services, clear emergency procedures and training. A small retailer should not need to invent its own crisis-payment policy. Essential merchants should have a tested plan for accepting cash, making change and communicating temporary limits without creating unnecessary confusion.

Households should not be left with the impression that resilience is someone else’s job. A modest cash reserve, a physical payment card, verified bank-contact details and an understanding of how to report a scam provide a practical buffer. The purpose is not to stockpile or to predict collapse. It is to avoid the avoidable problem of being unable to buy food, fuel or medicine because a phone is flat, a network is congested or a bank has paused an account for review. Resilience works best when public systems, businesses and households each maintain a small part of it.

The conversation should also remain honest about cash’s limits. It is vulnerable to theft and loss. It is awkward for remote sales. It does not give a buyer a built-in chargeback process. It can be used to conceal wrongdoing, particularly at larger values or within criminal schemes. Those limits justify proportionate security, targeted enforcement and modern digital alternatives. They do not justify making every lawful consumer transaction traceable, device-dependent and subject to automated permission. Public policy should reduce abuse without treating normal cash use as a problem to be designed away.

The importance of cash will therefore not be decided by whether it remains the most popular way to pay. It will be decided by whether it remains a usable way to pay when other methods fail or do not fit the circumstances. That is a different test. A person may tap a phone every day and still need notes when travelling, during an outage or while resolving an account issue. A merchant may prefer digital records and still need cash during a terminal failure. A country may lead in instant payments and still depend on physical currency in a serious disruption. Usage share alone does not measure fallback value.

This distinction is especially relevant in the euro area. The European Central Bank’s cash strategy expressly aims to keep euro cash widely available, accessible and accepted, while the European Commission’s legal-tender proposal is intended to protect access and acceptance in practice. The ECB’s work on the digital euro is framed as a complement to cash rather than a replacement. These policy positions are not a retreat from digital progress. They recognise that public money must remain usable across more than one technological setting.

The case for payment pluralism becomes stronger, not weaker, as artificial intelligence becomes more capable. AI can identify scam patterns, improve fraud controls and provide useful assistance. It can also make systems harder to understand, concentrate risk in common providers and create convincing new forms of impersonation. European and international authorities already warn that AI-supported phishing and deepfake-enabled fraud are part of the threat environment, while financial regulators are examining the governance risks of AI in finance. A payment system should respond by widening its safeguards, not by removing its independent fallback.

Cash is that fallback in its most practical form. It does not require a person to be identified by a device, approved by a behavioural score or connected to a live network at the moment of exchange. It cannot replace card payments, instant transfers or online banking, and it should not be asked to. Its role is to ensure that the ability to buy an ordinary good or service does not disappear when one digital route becomes unavailable, unsafe or unsuitable.

The policy bargain is straightforward. Society can embrace faster payments, stronger digital security, better offline tools and carefully governed AI. In return, it should keep physical cash obtainable and accepted, retain human routes through automated decisions, and avoid designing essential services around a single private technology stack. That bargain protects people who rely on cash today and people who may unexpectedly need it tomorrow.

The age of AI does not make cash less relevant because machines can process payments more quickly. It makes cash more valuable as a known limit on automation. A note does not learn from a customer, classify them, track them, require an update or decide that their behaviour is unusual. It simply remains available for a direct exchange. That quiet independence is precisely what a resilient payment system needs.

It also changes the political meaning of choice. Choice is not real merely because a consumer can download several wallets or select between competing cards. It is real when a person can still complete an ordinary payment after a phone is lost, a network is unavailable, an account is under review or a provider changes its rules. Cash gives that choice a form that does not depend on subscription terms, operating-system support or a successful automated identity check. It is not a substitute for fixing exclusion in digital services. It is the protection against exclusion becoming total while those services are fixed.

The most sensible future is therefore not cashless or anti-digital. It is deliberately mixed. Digital payments should become more secure, transparent and accessible; public authorities should enforce operational resilience and privacy rules; banks should offer prompt human help when automation fails; and merchants should retain the capacity to accept public money in physical form. This combination costs more than betting everything on one smooth digital channel. It also leaves fewer people stranded when smoothness breaks.

For policymakers, the final question is simple. When a resident needs to pay for an essential good under imperfect conditions, which options are genuinely available without needing permission from a model, a device or a distant platform? The answer should never be “none.” Cash helps ensure that it is not.

Questions readers ask about cash and AI

Author:
Jan Bielik
CEO & Founder of Webiano Digital & Marketing Agency

This article is an original analysis supported by the sources cited below

Study on the payment attitudes of consumers in the euro area
ECB data on the use of cash, cards and mobile payments at physical points of sale in 2024.

The Eurosystem cash strategy
ECB policy framework for keeping euro cash available, accessible and accepted.

The euro as legal tender
European Commission explanation of the proposal to safeguard euro-cash access and acceptance.

Digital euro package
European Commission legislative package for a possible digital euro and the legal-tender status of cash.

FAQs on the digital euro
ECB answers on the proposed digital euro’s legal-tender status, accessibility and relationship with cash.

Privacy and the digital euro
ECB material on privacy design, including the intended role of offline payments.

How the digital euro would work
ECB explanation of the project’s design principles, including the distinction between user-directed conditional payments and programmable money.

Keep calm and carry cash
ECB analysis of the role of physical cash during crises, including the April 2025 Iberian blackout.

Payments report 2025
Riksbank assessment of resilience, access and preparedness in a highly digital payment market.

Collaboration makes the payment system resilient
Riksbank update on offline card-payment arrangements for essential goods.

Instant payments
National Bank of Slovakia information on instant-payment adoption and SEPA transfer rules in Slovakia.

Status report on euro counterfeits recovered in 2025
National Bank of Slovakia data and guidance on counterfeit euro notes and coins recovered in Slovakia.

Project Hertha
BIS research into payment-transaction analytics for detecting financial-crime patterns while minimising data use.

Regulating AI in the financial sector
BIS Financial Stability Institute analysis of AI uses, governance and supervisory questions in finance.

Enhancing cross-border payments by addressing fraud
CPMI report on fraud risks in cross-border payments and the potential role of AI in detection.

Digital Operational Resilience Act
EU regulation governing digital operational resilience and ICT third-party risk in financial services.

ENISA Threat Landscape 2025
EU cybersecurity assessment covering AI-supported phishing and the changing threat environment.

Facing reality
Europol report on criminal uses of deepfake technology, including impersonation and CEO fraud.

The Global Findex Database 2025
World Bank evidence on digital connectivity, financial inclusion, account use and digital-safety gaps.

The rise of cyber events and digital fraud in the financial sector
IMF working paper on cyber incidents, digital fraud and payment-related scam risks.

Markets in Crypto-Assets Regulation
EUR-Lex summary of the EU’s regulatory framework for crypto-assets, including relevant stablecoin categories.