7 terms
AI alignment is the work of making AI systems pursue the goals, values and constraints their designers and users actually intend, even in new situations.
AI bias is a systematic error that makes an AI system produce unfair or skewed outcomes for certain groups, inputs or situations rather than random mistakes.
AI guardrails are the technical and policy controls placed around an AI model that check its inputs, outputs and actions against safety and business rules.
AI red teaming is the structured adversarial testing of AI models and systems to uncover harmful, insecure or unintended behavior before attackers or users do.
Explainable AI covers methods that make AI model decisions understandable to humans, supporting oversight, debugging, fairness review and accountability.
A model card is a structured document that describes a machine learning model's intended use, performance, limitations and risks for AI safety review.
Prompt injection is an attack in which crafted text makes a language model follow an attacker's instructions instead of those set by its developer or user.