WHAT THIS BILL REGULATES · 4 REQUIREMENT TYPES
How Is This Bill Enforced
Verbatim statutory text on the left; plain-language analysis and a per-section checklist on the right. Numbered markers cross-link to the matching checklist row.
This Act may be cited as the ''AI Foundation ModelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) Transparency Act of 2023''.
Establishes the short title of the Act as the "AI Foundation Model Transparency Act of 2023." This is a naming provision with no compliance obligation.
(1) With the increase in public access to artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1), there has been an increase in lawsuits and public concerns about copyright infringement, including in court cases such as the following: (A) Doe 1 v. GitHub, Inc., No. 22–cv–06823, 2023 WL 3449131, at *1 (N.D. Cal. May 11, 2023). (B) Amended Complaint, Getty Images, Inc. v. Stability AI, Ltd., No. 23–cv–00135 (D. Del. Mar. 29, 2023). (C) Andersen v. Stability AI Ltd., No. 23–cv–00201, 2023 WL 7132064, at *1 (N.D. Cal. Oct. 30, 2023).
(2) Public use of foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) has led to countless instances of the public being presented with inaccurate, imprecise, or biased information during inferenceInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5), based on limited training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6), limited model training mechanisms, or a lack of disclosures about the training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6) composition or foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) training procedures, including in facial recognition technology usage, artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1) inferencesInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5) relating to health, artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1) inferencesInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5) relating to loan granting and housing approval, and more.
(3) Transparency with respect to high-impact foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) has become increasingly necessary, including to assist copyright owners with enforcing their copyright protections and to promote consumer protection.
(4) While not compromising the intellectual property rights of those who develop and deploy foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), users should be equipped with the information necessary to enforce their copyright protections and to make informed decisions about such foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4).
Sets forth congressional findings regarding the increase in copyright infringement lawsuits related to AI training data, the prevalence of inaccurate or biased AI outputs stemming from limited training data transparency, and the need for foundation model transparency to support copyright enforcement and consumer protection. These are legislative purpose statements and create no compliance obligations.
(a)(1)–(2) 1 Not later than 9 months after the date of the enactment of this Act, the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) shall— (1) in accordance with section 553 of title 5, United States Code, promulgate regulations that establish standards specifying information to improve the transparency of foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) by covered entities with respect to training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6), model documentation, data collection in inferenceInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5), and operations of foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4); and (2) issue guidance to assist covered entities in complying with the standards established under paragraph (1).
This subsection imposes the bill's central regulatory mandate: the FTC must, within 9 months of enactment, promulgate regulations establishing transparency standards for foundation models and issue compliance guidance to covered entities. The rulemaking must follow notice-and-comment procedures under 5 U.S.C. § 553. The standards must address training data, model documentation, data collection during inference, and model operations.
(b) In establishing the standards and issuing the guidance required by subsection (a), the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) shall consult with the Director of the National Institute of Standards and Technology, the Director of the Office of Science and Technology Policy, the Register of Copyrights, and other relevant stakeholders, including standards bodies, covered entities, academia, technology experts, and advocates for civil rights and consumers.
Requires the FTC to consult with the Director of NIST, the Director of OSTP, the Register of Copyrights, and other relevant stakeholders — including standards bodies, covered entities, academia, technology experts, and civil rights and consumer advocates — when establishing the transparency standards and issuing guidance. This is a procedural requirement on the agency, not a compliance duty on covered entities.
(c)(1)–(2) 2 The standards established under subsection (a)(1) shall include requirements, with respect to a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), for— (1) what information specified under such subsection shall be submitted to the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) by the covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) that provides such model; and (2) what information specified under such subsection shall be made publicly available by the covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) that provides such model.
Requires that the FTC's transparency standards bifurcate covered-entity obligations into two channels: information that must be submitted to the Commission, and information that must be made publicly available by the covered entity. This creates the dual-track disclosure architecture — regulator submission plus public transparency — that underlies the bill's core compliance burden on covered entities.
(d)(1)–(5) 3 The standards established under subsection (a)(1) shall specify the form and manner in which certain information specified under such subsection, selected at the discretion of the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2), in consultation with the Director of the National Institute of Standards and Technology and the other actors described in subsection (b), shall be made publicly available by covered entities, including— (1) what information shall be made available on the website of a covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) that relates to any foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) provided by such covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3); (2) what information shall be displayed in a central location on a website hosted by the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2), which shall include, with respect to a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), information that is substantially similar to the information required under paragraph (1) to be made available on the website of the covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) that provides such model; (3) that a machine-readable format shall be used with respect to the information specified under paragraphs (1) and (2); (4) the URL at which the information specified under paragraph (2) shall be hosted by the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2); and (5) such additional specifications as the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) considers appropriate.
Specifies the structural requirements for how covered entities must make foundation model information publicly available. The FTC must define what information appears on the covered entity's own website, what information appears on a central FTC-hosted website, and that a machine-readable format must be used for both. The FTC must also specify the URL for its central registry. This subsection operationalizes the public transparency prong of the dual-track disclosure framework established in subsection (c).
(e) The standards established under subsection (a)(1) shall specify a process by which the information required under subsection (c)(1) shall be submitted to the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2).
Directs the FTC to specify a process for covered entities to submit the information required under subsection (c)(1) to the Commission. This is a procedural directive to the agency governing how regulator submissions are received.
(f)(1) 4 The sources of training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6) (including, as applicable, personal data collection and information necessary to assist copyright owners or data license holders with enforcing their copyright or data license protections) and whether and how data is collected and retained during inferenceInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5).
(f)(2) 4 A description of the size and composition of such training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6), including broad demographic information, language information, and other attribute information, while accounting for privacy.
(f)(3) 4 Information on data governance procedures, including how such training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6) was edited or filtered.
(f)(4) 4 How such training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6) was labeled, and information regarding how the validity of the labeling process was assessed.
(f)(5) 5 A description of the intended purposes and foreseen limitations or risks of the foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), an overview of past edits to such model, the version of such model, and the date of release of such model.
(f)(6) 6 A description of the efforts of the covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) to align the foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) and the transparency of such model with— (A) the AI Risk Management Framework (or any successor framework) of the National Institute of Standards and Technology; or (B) a similar Federal Government-approved consensus technical standard.
(f)(7) 7 Performance under evaluation, either self-driven or through audit, on public or industry standard benchmarks, including what precautions the foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) takes to answer or respond to situations with higher levels of risk of providing inaccurate or harmful information, including, if such model responds to such questions, relating to the following: (A) Medical, health, or healthcare questions. (B) Biological or chemical synthesis. (C) Cybersecurity. (D) Elections. (E) Policing, including predictive policing. (F) Financial loan decisions. (G) Education. (H) Employment or hiring decisions. (I) Public services. (J) Information relating to vulnerable populations, including children and protected classes.
(f)(8) 5 Information on the computational power used to train and operate the foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4).
(f)(9) Any other information determined necessary by the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2), in consultation with the Director of the National Institute of Standards and Technology, to improve transparency of foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4).
Enumerates the categories of information the FTC must consider including in the transparency standards. The list covers training data sources (including copyright- and license-relevant information), data size and composition, data governance procedures, data labeling methods, model purposes and limitations, NIST AI RMF alignment, benchmark performance in high-risk domains (healthcare, biosecurity, cybersecurity, elections, policing, finance, education, employment, public services, and vulnerable populations), computational power, and any other information the Commission deems necessary. This subsection is the substantive backbone of the bill's disclosure mandate — it tells the FTC what topics the covered-entity disclosure must address.
(g)(1)–(2) In establishing the standards and issuing the guidance required by subsection (a), the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) shall consider whether to include alternative provisions for— (1) open-source foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4); or (2) foundation modelsFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) that are a derivation of or built upon another foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), having been retrained or adapted from such other foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) to any extent.
Directs the FTC to consider whether to establish alternative or modified transparency requirements for two categories of foundation models: open-source models and derivative models that are built upon, retrained from, or adapted from another foundation model. This is a directive to the agency to consider differentiated treatment, not a compliance obligation on covered entities.
(h) The regulations required by subsection (a)(1) shall apply beginning on the date that is 90 days after the date on which the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) promulgates such regulations.
Establishes the operative date for the regulations: 90 days after the FTC promulgates them. This staged effective date means covered-entity obligations do not begin until 9 months (rulemaking deadline) plus 90 days after enactment at the earliest.
(i) Not later than 2 years after the date on which the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) promulgates the regulations required by subsection (a)(1), and not less often than annually thereafter, the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2), in consultation with the Director of the National Institute of Standards and Technology, shall assess the standards established by such regulations and update such regulations so as to incorporate appropriate updates (if any) to such standards.
Requires the FTC, in consultation with NIST, to assess and update the transparency standards not later than 2 years after initial promulgation, and at least annually thereafter. This creates a periodic review cadence ensuring the standards evolve with the technology.
(j)(1) A violation of a regulation promulgated under subsection (a)(1) shall be treated as a violation of a regulation under section 18(a)(1)(B) of the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) Act (15 U.S.C. 57a(a)(1)(B)) regarding unfair or deceptive acts or practices.
(j)(2)(A)–(B) Except as provided in subsection (m)(3)(C)— (A) the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) shall enforce the regulations promulgated under subsection (a)(1) in the same manner, by the same means, and with the same jurisdiction, powers, and duties as though all applicable terms and provisions of the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) Act (15 U.S.C. 41 et seq.) were incorporated into and made a part of this section; and (B) any covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3) that violates a regulation promulgated under subsection (a)(1) shall be subject to the penalties and entitled to the privileges and immunities provided in the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) Act.
Establishes the enforcement mechanism. Violations of the transparency regulations are treated as violations of an FTC Act section 18(a)(1)(B) regulation regarding unfair or deceptive acts or practices, giving the FTC full enforcement authority including the penalties available under the FTC Act. The FTC's jurisdiction is expanded beyond its normal scope to include common carriers subject to the Communications Act and nonprofit organizations.
(k) Not later than 2 years after the date of the enactment of this Act, the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) shall submit to the Committee on Energy and Commerce and the Committee on Science, Space, and Technology of the House of Representatives and the Committee on Commerce, Science, and Transportation of the Senate a report on the establishment, implementation, and enforcement of the standards required by subsection (a)(1).
Requires the FTC to submit a report to designated House and Senate committees within 2 years of enactment on the establishment, implementation, and enforcement of the transparency standards. This is an agency reporting obligation to Congress, not a compliance duty on covered entities.
(l)(1)–(2) There are authorized to be appropriated to the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) to carry out this section— (1) $10,000,000 for fiscal year 2025; and (2) $3,000,000 for each fiscal year thereafter.
Authorizes $10 million for fiscal year 2025 and $3 million annually thereafter for the FTC to carry out the Act. This is a funding provision with no compliance obligation on covered entities.
(m)(1) The term ''artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1)'' has the meaning given such term in section 5002 of the National Artificial IntelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1) Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).
(m)(2) The term ''CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2)'' means the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2).
(m)(3)(A)–(C) The term ''covered entityCovered entityThe term "covered entity" means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation model which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation model which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. The persons, partnerships, and corporations described in subparagraph (C) are— (i) any person, partnership, or corporation over which the Commission has jurisdiction under section 5(a)(2) of the Federal Trade Commission Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade Commission Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the Commission— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.Sec. 3(m)(3)'' means any person, partnership, or corporation described in subparagraph (C) that provides— (i) use of or services from a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) which generates, in aggregate, over 100,000 monthly output instances (whether text, images, video, audio, or other modality), including output instances generated from use by users of second party entities that use such model; or (ii) use of or services from a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) which has, in aggregate, over 30,000 monthly users, including users of second party entities that use such model. (B) The CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2), in consultation with the Director of the National Institute of Standards and Technology and the Director of the Office of Science and Technology Policy, may, by regulation promulgated in accordance with section 553 of title 5, United States Code, update the number of monthly output instances for purposes of subparagraph (A)(i) or the number of monthly users for purposes of subparagraph (A)(ii), as the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) considers appropriate. (C) The persons, partnerships, and corporations described in this subparagraph are— (i) any person, partnership, or corporation over which the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) has jurisdiction under section 5(a)(2) of the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) Act (15 U.S.C. 45(a)(2)); and (ii) notwithstanding section 4, 5(a)(2), or 6 of the Federal Trade CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2) Act (15 U.S.C. 44; 45(a)(2); 46) or any jurisdictional limitation of the CommissionCommissionThe term "Commission" means the Federal Trade Commission.Sec. 3(m)(2)— (I) any common carrier subject to the Communications Act of 1934 (47 U.S.C. 151 et seq.) and all Acts amendatory thereof and supplementary thereto; and (II) any organization not organized to carry on business for its own profit or that of its members.
(m)(4)(A)–(B) The term ''foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4)'' means an artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1) model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. (B) The term ''foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4)'' includes an artificial intelligenceArtificial intelligenceThe term "artificial intelligence" has the meaning given such term in section 5002 of the National Artificial Intelligence Initiative Act of 2020 (15 U.S.C. 9401; enacted as division E of the William M. (Mac) Thornberry National Defense Authorization Act for Fiscal Year 2021 (Public Law 116–283)).Sec. 3(m)(1) model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.
(m)(5) The term ''inferenceInferenceThe term "inference" means, with respect to a foundation model, when such foundation model is operated by a user to produce a result.Sec. 3(m)(5)'' means, with respect to a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), when such foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) is operated by a user to produce a result.
(m)(6) The term ''training dataTraining dataThe term "training data" means, with respect to a foundation model, the data on which such foundation model was trained.Sec. 3(m)(6)'' means, with respect to a foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4), the data on which such foundation modelFoundation modelThe term "foundation model" means an artificial intelligence model that— (i) is trained on broad data; (ii) generally uses self-supervision; (iii) generally contains at least 1,000,000,000 parameters; (iv) is applicable across a wide range of contexts; and (v) exhibits, or could be easily modified to exhibit, high levels of performance at tasks that could pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters. The term "foundation model" includes an artificial intelligence model otherwise described in subparagraph (A) even if such model is provided to users with technical safeguards that attempt to prevent users from taking advantage of any relevant unsafe capabilities.Sec. 3(m)(4) was trained.
Defines six key terms used throughout the Act: artificial intelligence (by cross-reference to the National AI Initiative Act of 2020), Commission (the FTC), covered entity (persons providing foundation models exceeding defined usage thresholds, including common carriers and nonprofits), foundation model (broad-data, self-supervised, ≥1B parameter models applicable across wide contexts and capable of posing serious risks), inference, and training data. The covered entity definition sets the scope of the Act's obligations through a dual threshold: either 100,000 monthly output instances or 30,000 monthly users. The foundation model definition includes models with technical safeguards.