Definitions

In this Act: (1) Artificial intelligence; artificial intelligence system The terms artificial intelligence and artificial intelligence system has the meaning given the term artificial intelligence in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401). (2) Artificial intelligence agent The term artificial intelligence agent — (A) means an artificial intelligence system or process that, given an objective or instruction— (i) determines the action or sequence of actions to be taken to accomplish that objective; and (ii) is capable of executing such actions directly on information systems, data, or external services; and (B) does not include a system or process that solely generates informational or advisory output for a human operator to act upon. (3) Artificial intelligence capabilities or risk assessment The term artificial intelligence capabilities or risk assessment means information regarding an assessment performed by a developer of an artificial intelligence system evaluating the capabilities of, or risks posed by, such system, including an assessment of the potential of the artificial intelligence system— (A) to materially assist in the design, development, acquisition, or use of a chemical, biological, radiological, or nuclear weapon; (B) to materially assist in the design, development, or production of munitions or other weapons; (C) to materially assist in the unlawful manufacture, synthesis, or distribution of a controlled substance; (D) to evade the control of, or act outside the intended instructions of, its developer or operator; (E) to facilitate a cybersecurity threat, including through the discovery or exploitation of a security vulnerability; (F) to be subject to the unauthorized exfiltration of the weights of the artificial intelligence system, or unauthorized, deliberate, malicious modification of such weights; or (G) to be transformed, stolen, reverse-engineered, or otherwise manipulated by an unauthorized user or users acting outside of the terms of service governing access to the artificial intelligence system. (4) Artificial intelligence flaw The term artificial intelligence flaw means a recurring or reproducible characteristic, behavior, or failure mode of an artificial intelligence system that causes, or materially increases the risk of, an artificial intelligence safety incident absent any intentional act of a user, including a characteristic, behavior, or failure mode that may manifest across multiple systems or providers. (5) Artificial intelligence safety incident The term artificial intelligence safety incident means an event that materially increases the risk that operation of an artificial intelligence system leads to a state in which human life, health, property, or the environment is endangered. (6) Artificial intelligence security incident The term artificial intelligence security incident means an event that materially increases— (A) the risk that operation of an artificial intelligence system occurs in a way that enables the unauthorized extraction of information about the behavior or characteristics of the system by an unauthorized party; or (B) the ability to manipulate an artificial intelligence system in order to subvert the confidentiality, integrity, or availability of the system or adjacent system. (7) Artificial intelligence security vulnerability The term artificial intelligence security vulnerability means a weakness in an artificial intelligence system that could be exploited by a third party to subvert, without authorization, the confidentiality, integrity, or availability of the system, including through techniques such as— (A) data poisoning; (B) evasion attacks; (C) privacy-based attacks; (D) model theft or extraction attacks; (E) attacks designed to circumvent or degrade the safety, alignment, or access control mechanisms of an artificial intelligence system; and (F) adversarial machine learning attacks as described in National Institute of Standards and Technology Trustworthy and Responsible Artificial Intelligence 100–2e2025 (relating to Adversarial Machine Learning), or successor publication. (8) Board The term Board means the Artificial Intelligence Safety Board established under section 3. (9) Critical infrastructure The term critical infrastructure has the meaning provided in section 1016(e) of the USA Patriot Act of 2001 (42 U.S.C. 5195c(e)). (10) Developer The term developer means a developer of a frontier artificial intelligence model. (11) Frontier artificial intelligence model The term frontier artificial intelligence model means an artificial intelligence model, or system combining multiple artificial intelligence models, that exhibits or could be modified to exhibit high levels of performance at tasks that pose a serious risk to national security, national economic security, or public health or safety. (12) Institute The term Institute means the National Institute of Standards and Technology. (13) Secretary The term Secretary means the Secretary of Commerce.