AI Glossary

An independent review of an AI system in terms of regulatory compliance, data quality, risks and impacts on the persons concerned.

The use of AI for diagnostics, decision support or the processing of health data; often falling under both the medical devices regime and that for high-risk systems.

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.

The contractual and statutory definition of which entity bears which obligations and risks; particularly crucial for high-risk systems and the integration of third-party models.

A reference to EU harmonization legislation; AI systems embedded in regulated products (e.g., medical devices, machinery) are classified as high-risk under it, with a longer transition period.

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 company developing the Claude family of models, with an emphasis on safety and the Constitutional AI approach; a party to the Bartz case and its settlement, and a participant in discussions about GPAI models with systemic risk.

The provision laying down the specific obligations of providers of GPAI models with systemic risk, in particular risk assessment and mitigation, incident reporting and ensuring security.

The provision setting out the rules under which an AI system is considered high-risk, including exemptions for systems that do not pose a significant risk of harm.