Definitions

In this Act: (1) AI red-teaming The term AI red-teaming means a structured testing effort to find flaws and vulnerabilities of an AI system, often performed— (A) in a controlled environment and in collaboration with the developers of the AI system; and (B) by dedicated teams that adopt adversarial methods to identify any flaws or vulnerabilities of the AI system, which may include harmful or discriminatory outputs, unforeseen or undesirable system behaviors, limitations, and potential risks associated with misuse. (2) Artificial intelligence model The term artificial intelligence model means the part of an artificial intelligence system that can be adjusted to modify how the system behaves during inference. (3) Artificial intelligence system The term 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). (4) Cloud provider The term cloud provider means an entity that offers on-demand network access to a shared pool of configurable computing resources, such as networks, servers, storage, and applications that can be rapidly provisioned and released with minimal management effort. (5) Commission The term Commission means the Federal Trade Commission. (6) Covered deployer The term covered deployer means an entity that operates a covered system which— (A) for the majority of the months in any 1-year period, has, or is expected to have, not fewer than 1,000,000 unique monthly users in the United States; or (B) allows minors to create accounts or profiles and has, or is expected to have, not fewer than 100,000 unique monthly users in the United States. (7) Covered model The term covered model means a generative artificial intelligence model, including an open weight model or an open source model, that— (A) either— (i) is accessible to the public through a website, software application, an application programming interface, or a licensure through a cloud provider, including as part of a broader AI system; or (ii) the model weights or architecture of which are made available to the public through the website of a developer, download partners, cloud partners, or software development platform, including in a situation in which accessing such model weights or architecture requires a user to create an account; and (B) is deployed in covered systems, whether deployed by the developer or another deployer, that— (i) together have, or are expected to have, not fewer than 1,000,000 unique monthly users in the United States for a majority of the months in any 1-year period; or (ii) allow minors to create accounts or profiles and that together have, or are expected to have, not fewer than 100,000 unique monthly users in the United States for a majority of the months in any 1-year period. (8) Covered platform The term covered platform means a website or software application used to interface with a covered system. (9) Covered system The term covered system means a covered model or collection of covered models, together with any software component or infrastructure that— (A) are deployed as an integrated unit; and (B) operate in a coordinated manner to— (i) handle different tasks or modalities; or (ii) materially augment or enhance the capabilities of the covered model. (10) Developer The term developer — (A) means an entity that initially builds, designs, codes, produces, pre-trains, or owns an artificial intelligence system for internal use or for use by a third party; and (B) does not include an entity that is solely— (i) a deployer of the artificial intelligence system; or (ii) a substantial modifier of the artificial intelligence system. (11) Generative artificial intelligence model The term generative artificial intelligence model means an artificial intelligence model that— (A) emulates the structure and characteristics of input data in order to generate derived synthetic data, including images, videos, audio, text, code, and other digital content; (B) is trained on broad data; (C) generally uses self-supervision; (D) is designed for generality of output; and (E) is generally applicable across a wide range of contexts, adaptable to a wide range of tasks, or can issue a wide range of outputs in response to inferences. (12) Machine learning The term machine learning has the meaning given the term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401). (13) Minor The term minor means an individual that has not attained 18 years of age. (14) Model inference The term model inference means the process of running data points through an AI system to receive an output, such as a number or text. (15) Model weight The term model weight means a numerical parameter within an artificial intelligence model that helps determine the outputs of the model in response to inputs. (16) Open weight model The term open weight model means a machine learning model, the weights of which are made publicly available, allowing for access, inspection, modification, and redistribution by any third party. (17) Open source model The term open source model means a machine learning model, the weight and model architecture, training code, and methodology of which are made publicly available, allowing for a third party to fully reproduce, audit, and modify the model for custom applications. (18) Post training and inference enhancements The term post training and inference enhancements means a process of adjusting or supplementing a pre-trained artificial intelligence model to improve performance on specific tasks, better meet desired performance characteristics, or adhere to safety constraints. (19) Senior The term senior means an individual that has attained 65 years of age. (20) Substantial modifier The term substantial modifier means an entity that significantly alters (as defined by the Commission) the behavior or capabilities of a covered model through post training and inference enhancements. (21) Synthetic data The term synthetic data means data that is artificially manufactured, typically by a computer program. (22) Vulnerable user The term vulnerable user means a minor, a senior, an individual with a disability, or an individual experiencing a mental health crisis.