What Is AI Assurance?
AI assurance is the continuing work of collecting evidence that an AI system performs as intended and that important risks are understood and controlled.
Don't Just Read it – Explore It
AI assurance is the continuing work of collecting evidence that an AI system performs as intended and that important risks are understood and controlled.
An AI risk register is a maintained record of identified risks, their likelihood and impact, responsible owners, controls, status, and planned treatment.
An algorithmic audit systematically evaluates an automated system, its development process, and its outcomes against defined criteria and evidence.
AI data provenance is the recorded history of where data originated, who or what changed it, and how it moved through a system.
An AI impact assessment examines how an automated system may affect people, organizations, rights, services, and communities before and during deployment.
A dataset card is a concise, structured description of a dataset’s origins, contents, intended uses, collection choices, limitations, and maintenance history.
A datasheet for a dataset records how data was motivated, composed, collected, processed, distributed, maintained, and intended to be used.
An AI system card documents how an AI product was evaluated, what risks were considered, which safeguards are used, and where important limitations remain.
LLM-as-a-judge uses a language model to score, rank, or critique outputs according to instructions or a rubric.
Pairwise evaluation asks a judge to choose between two model outputs for the same task instead of assigning each response an absolute score.
Representation engineering identifies and manipulates high-level directions in a model’s internal activations to analyze or influence behavior.
An AI evaluation harness runs models through standardized tasks, prompts, scoring rules, and reporting so results can be reproduced and compared.