AI Glossary

A continuously maintained overview of all AI systems used in an organization, enabling oversight to be managed according to the risk level of individual systems.

A continuously maintained inventory of all AI systems used within an organization, enabling oversight to be managed according to the level of risk.

The ability to continuously monitor the behavior of a deployed system (inputs, outputs, errors, costs) for the purposes of debugging, auditing, and risk management.

A list of areas in which standalone AI systems are considered high-risk, such as recruitment, credit scoring, education, law enforcement, migration and border management, and critical infrastructure.

The distinction between evaluating what a model can do (its capabilities) and evaluating whether its use is safe; both form part of managing systemic risks.

System design in which legal and security requirements, defined purposes of use, risk management and ongoing auditability are built in from the development stage onwards, so that the agent remains compliant with regulation.

An AI system falling into the high-risk category under the AI Act, subject to strict requirements on risk management, data quality, documentation, transparency and human oversight.

A set of processes through which a provider continuously identifies, assesses, monitors and mitigates the risks associated with an AI system.