AI Glossary
A
Measures ensuring the protection of an AI system against unauthorized interference, attacks and misuse.
A review of AI systems, data and processes carried out, for example, in the context of investments or acquisitions.
A set of ethical principles relating to the development and use of artificial intelligence.
The system of rules, roles and processes by which an organization governs the development, deployment and use of AI in compliance with the law and with ethical standards.
A situation in which a model produces a convincing-seeming yet false or fabricated output that is not supported by data or facts.
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.
The deployment of AI in legal research, document analysis and decision support; in some contexts classified as high-risk or subject to specific regulation.
The use of AI by public authorities, for example to sort submissions or support decision-making; particularly sensitive from the perspective of transparency and fundamental rights.
An event in which an AI system causes damage, a failure or another undesirable consequence that may require reporting.
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 proposed EU regulation intended to make it easier to pursue claims for damages caused by AI; its fate has been the subject of debate, and the Commission has been reconsidering it as part of its simplification efforts.
A sufficient level of knowledge and skills to use AI responsibly and to understand both its opportunities and its risks.
The person responsible for coordinating the AI agenda within an organization.
A limited deployment of AI for the purpose of evaluating its functioning and benefits.
The process of selecting and acquiring AI solutions, including an assessment of the related risks and legal compliance requirements.
Targeted probing of a model by a team that deliberately tries to induce harmful or dangerous behavior, so that weaknesses can be identified and removed before deployment.
A continuously maintained inventory of all AI systems used within an organization, enabling oversight to be managed according to the level of risk.
Classifying AI systems into tiers according to their potential impact on individuals, organizations, and society in order to select proportionate oversight measures.
An international gathering of representatives of states, companies, and experts dedicated to the risks of the most advanced models and to coordinating approaches to their safety.
An entity providing an AI solution or service to a customer; its obligations and liability are usually defined by contract as well as by legal regulations.
A group of people responsible for the strategic governance of the use of AI within an organization.
A legal construction under which, for the purposes of market surveillance, an AI system is treated as a product, so that the existing supervisory mechanisms apply.
An event or failure of an AI system that led or could have led to a serious threat to the health, safety or fundamental rights of persons.
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 natural or legal person that develops an AI system or has one developed and places it on the market or puts it into service under its own name or trademark.
The system's resilience to errors, manipulation, attacks, or unexpected situations.
A requirement that users be informed that they are communicating or working with artificial intelligence.
The sequence of actors from the development of a model through its integration to its deployment; the AI Act distributes obligations among the individual links of the chain.
A situation in which the pricing or trading algorithms of several competitors effectively coordinate market behavior without any express agreement between them, raising competition law questions.
The effort to ensure that AI systems do not create unjustified differences between groups of people; it encompasses various, sometimes mutually incompatible, definitions of fairness.