{"check":null,"uid":"c1129c1c21271e94","title":"AI Systems Transparency Act","title_generated":false,"country":"США","organ":"Конгресс США","kind":"law","kind_name":"Законодательство","lang":"en","date":"2026-09-23","summary":"Закон требует раскрытия определённой информации об искусственных интеллектуальных системах, включая модели ИИ и платформы их использования. Он устанавливает критерии определения рисков безопасности, накладывает обязательства на разработчиков и операторов таких систем, касающиеся версии моделей, дизайна и обязательных раскрытых сведений. Нарушение закона рассматривается как нарушение правил Федеральной торговой комиссии США.","snippet":"","topics":["Искусственный интеллект","Регулирование контента","Цифровые платформы и сервисы"],"status":"ok","error":"","text_len":24840,"versions":1,"url":"https://www.congress.gov/bill/119th-congress/senate-bill/5471","first_seen":"2026-10-07","last_checked":"2026-11-16 05:19","relevance":"hit","score":41,"query":"","source_key":"congress_us","verdict":{"relevance":"hit","score":41,"topics":["Искусственный интеллект","Регулирование контента","Цифровые платформы и сервисы"],"need_body":3,"authorities":[{"kind":"орган","name":"Federal Trade Commission","topic":"Цифровые платформы и сервисы"}],"evidence":[{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":228,"ctx":"9th congress 2d session s. 5471  to require disclosure of certain information relating to artificial intelligence systems, and for other purposes.  _______________________________________________________","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":743,"ctx":"_______________________  a bill  to require disclosure of certain information relating to artificial intelligence systems, and for other purposes.  be it enacted by the senate and house of representative","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1555,"ctx":"ndesirable system behaviors, limitations, and potential risks associated with misuse. (2) artificial intelligence model.--the term ``artificial intelligence model'' means the part of an artificial intell","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1598,"ctx":"nd 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 modif","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1651,"ctx":"cial 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) artifi","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1758,"ctx":"igence 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 ``art","zone":"текст","weight":1},{"topic":"Регулирование контента","term":"child sexual abuse material","weak":false,"pos":13406,"ctx":"taken to ensure the safety of minors in training data, such as through data filtering of child sexual abuse material; (iv) an overview of privacy standards pertaining to the management, use, storage, and po","zone":"текст","weight":1},{"topic":"Регулирование контента","term":"child sexual abuse material","weak":false,"pos":13809,"ctx":"bility and efficacy of vulnerable user safeguards (including preventing the generation of child sexual abuse material and non-consensual intimate imagery), as well as the resiliency and scope of such feature","zone":"текст","weight":1},{"topic":"Регулирование контента","term":"non-consensual intimate","weak":false,"pos":13841,"ctx":"e user safeguards (including preventing the generation of child sexual abuse material and non-consensual intimate imagery), as well as the resiliency and scope of such features; (vi) an overview of user and comp","zone":"текст","weight":1},{"topic":"Регулирование контента","term":"child sexual abuse material","weak":false,"pos":20664,"ctx":"view of information pertaining to attempts by users to utilize covered models to generate child sexual abuse material or non- consensual intimate imagery; and (iv) an overview of known system capabilities fo","zone":"текст","weight":1},{"topic":"Регулирование контента","term":"intimate image","weak":true,"pos":20711,"ctx":"sers to utilize covered models to generate child sexual abuse material or non- consensual intimate imagery; and (iv) an overview of known system capabilities for the utilization of the covered mod","zone":"текст","weight":0},{"topic":"Цифровые платформы и сервисы","term":"Federal Trade Commission","weak":false,"pos":2356,"ctx":"leased 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 cov","zone":"орган","weight":3}],"dropped":[{"topic":"Электронная коммерция и платежи","term":"digital content","weak":false,"pos":5121,"ctx":"o 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 g","why":"одиночное упоминание (нужно 3)"},{"topic":"Персональные данные","term":"privacy","weak":true,"pos":13071,"ctx":"loyment testing and model development; (ii) an overview of measures taken to protect user privacy (including the privacy of minors) in training data, such as through data filtering or ano","why":"только многозначные термины"},{"topic":"Персональные данные","term":"privacy","weak":true,"pos":13094,"ctx":"el development; (ii) an overview of measures taken to protect user privacy (including the privacy of minors) in training data, such as through data filtering or anonymization methods, and","why":"только многозначные термины"},{"topic":"Персональные данные","term":"privacy","weak":true,"pos":13223,"ctx":"ch as through data filtering or anonymization methods, and measures taken to protect user privacy (including the privacy of minors) on covered platforms; (iii) an overview of measures tak","why":"только многозначные термины"},{"topic":"Персональные данные","term":"privacy","weak":true,"pos":13246,"ctx":"ering or anonymization methods, and measures taken to protect user privacy (including the privacy of minors) on covered platforms; (iii) an overview of measures taken to ensure the safety","why":"только многозначные термины"},{"topic":"Персональные данные","term":"privacy","weak":true,"pos":13455,"ctx":"data, such as through data filtering of child sexual abuse material; (iv) an overview of privacy standards pertaining to the management, use, storage, and potential sale of user data (in","why":"только многозначные термины"},{"topic":"Онлайн-реклама","term":"targeted advertising","weak":false,"pos":13621,"ctx":"sale of user data (including the data of minors), including whether user data is used for targeted advertising, user safety, internal or external research, and model training; (v) the durability and e","why":"одиночное упоминание (нужно 3)"},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":16397,"ctx":"ct the covered model and training data from attack or misuse, including-- (i) traditional cybersecurity protocols, such as access controls and multi-factor authentication; and (ii) protections","why":"одиночное упоминание (нужно 3)"},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":21240,"ctx":"to disclose any information that, as determined by the commission-- (a) would compromise cybersecurity practices or protections or safety practices or protections of the developer; or (b) coul","why":"одиночное упоминание (нужно 3)"}]},"last_changed":"2026-10-07","meta":{"congress":"119","billType":"S","number":"5471","policyArea":"Commerce","subjects":[],"latestAction":"2026-09-23 Read twice and referred to the Committee on Commerce, Science, and Transportation.","textVersion":"is"},"source_url":"https://www.congress.gov/bill/119th-congress/senate-bill/5471","text":"FULL TEXT\n[Congressional Bills 119th Congress]\n[From the U.S. Government Publishing Office]\n[S. 5471 Introduced in Senate (IS)]\n\n<DOC>\n\n119th CONGRESS\n2d Session\nS. 5471\n\nTo require disclosure of certain information relating to artificial\nintelligence systems, and for other purposes.\n\n_______________________________________________________________________\n\nIN THE SENATE OF THE UNITED STATES\n\nSeptember 23, 2026\n\nMr. Coons (for himself, Mrs. Britt, Mr. Schatz, and Mr. Lankford)\nintroduced the following bill; which was read twice and referred to the\nCommittee on Commerce, Science, and Transportation\n\n_______________________________________________________________________\n\nA BILL\n\nTo require disclosure of certain information relating to artificial\nintelligence systems, and for other purposes.\n\nBe it enacted by the Senate and House of Representatives of the\nUnited States of America in Congress assembled,\n\nSECTION 1. SHORT TITLE.\n\nThis Act may be cited as the ``AI Systems Transparency Act''.\n\nSEC. 2. DEFINITIONS.\n\nIn this Act:\n(1) AI red-teaming.--The term ``AI red-teaming'' means a\nstructured testing effort to find flaws and vulnerabilities of\nan AI system, often performed--\n(A) in a controlled environment and in\ncollaboration with the developers of the AI system; and\n(B) by dedicated teams that adopt adversarial\nmethods to identify any flaws or vulnerabilities of the\nAI system, which may include harmful or discriminatory\noutputs, unforeseen or undesirable system behaviors,\nlimitations, and potential risks associated with\nmisuse.\n(2) Artificial intelligence model.--The term ``artificial\nintelligence model'' means the part of an artificial\nintelligence system that can be adjusted to modify how the\nsystem behaves during inference.\n(3) Artificial intelligence system.--The term ``artificial\nintelligence system'' has the meaning given the term\n``artificial intelligence'' in section 5002 of the National\nArtificial Intelligence Initiative Act of 2020 (15 U.S.C.\n9401).\n(4) Cloud provider.--The term ``cloud provider'' means an\nentity that offers on-demand network access to a shared pool of\nconfigurable computing resources, such as networks, servers,\nstorage, and applications that can be rapidly provisioned and\nreleased with minimal management effort.\n(5) Commission.--The term ``Commission'' means the Federal\nTrade Commission.\n(6) Covered deployer.--The term ``covered deployer'' means\nan entity that operates a covered system which--\n(A) for the majority of the months in any 1-year\nperiod, has, or is expected to have, not fewer than\n1,000,000 unique monthly users in the United States; or\n(B) allows minors to create accounts or profiles\nand has, or is expected to have, not fewer than 100,000\nunique monthly users in the United States.\n(7) Covered model.--The term ``covered model'' means a\ngenerative artificial intelligence model, including an open\nweight model or an open source model, that--\n(A) either--\n(i) is accessible to the public through a\nwebsite, software application, an application\nprogramming interface, or a licensure through a\ncloud provider, including as part of a broader\nAI system; or\n(ii) the model weights or architecture of\nwhich are made available to the public through\nthe website of a developer, download partners,\ncloud partners, or software development\nplatform, including in a situation in which\naccessing such model weights or architecture\nrequires a user to create an account; and\n(B) is deployed in covered systems, whether\ndeployed by the developer or another deployer, that--\n(i) together have, or are expected to have,\nnot fewer than 1,000,000 unique monthly users\nin the United States for a majority of the\nmonths in any 1-year period; or\n(ii) allow minors to create accounts or\nprofiles and that together have, or are\nexpected to have, not fewer than 100,000 unique\nmonthly users in the United States for a\nmajority of the months in any 1-year period.\n(8) Covered platform.--The term ``covered platform'' means\na website or software application used to interface with a\ncovered system.\n(9) Covered system.--The term ``covered system'' means a\ncovered model or collection of covered models, together with\nany software component or infrastructure that--\n(A) are deployed as an integrated unit; and\n(B) operate in a coordinated manner to--\n(i) handle different tasks or modalities;\nor\n(ii) materially augment or enhance the\ncapabilities of the covered model.\n(10) Developer.--The term ``developer''--\n(A) means an entity that initially builds, designs,\ncodes, produces, pre-trains, or owns an artificial\nintelligence system for internal use or for use by a\nthird party; and\n(B) does not include an entity that is solely--\n(i) a deployer of the artificial\nintelligence system; or\n(ii) a substantial modifier of the\nartificial intelligence system.\n(11) Generative artificial intelligence model.--The term\n``generative artificial intelligence model'' means an\nartificial intelligence model that--\n(A) emulates the structure and characteristics of\ninput data in order to generate derived synthetic data,\nincluding images, videos, audio, text, code, and other\ndigital content;\n(B) is trained on broad data;\n(C) generally uses self-supervision;\n(D) is designed for generality of output; and\n(E) is generally applicable across a wide range of\ncontexts, adaptable to a wide range of tasks, or can\nissue a wide range of outputs in response to\ninferences.\n(12) Machine learning.--The term ``machine learning'' has\nthe meaning given the term in section 5002 of the National\nArtificial Intelligence Initiative Act of 2020 (15 U.S.C.\n9401).\n(13) Minor.--The term ``minor'' means an individual that\nhas not attained 18 years of age.\n(14) Model inference.--The term ``model inference'' means\nthe process of running data points through an AI system to\nreceive an output, such as a number or text.\n(15) Model weight.--The term ``model weight'' means a\nnumerical parameter within an artificial intelligence model\nthat helps determine the outputs of the model in response to\ninputs.\n(16) Open weight model.--The term ``open weight model''\nmeans a machine learning model, the weights of which are made\npublicly available, allowing for access, inspection,\nmodification, and redistribution by any third party.\n(17) Open source model.--The term ``open source model''\nmeans a machine learning model, the weight and model\narchitecture, training code, and methodology of which are made\npublicly available, allowing for a third party to fully\nreproduce, audit, and modify the model for custom applications.\n(18) Post training and inference enhancements.--The term\n``post training and inference enhancements'' means a process of\nadjusting or supplementing a pre-trained artificial\nintelligence model to improve performance on specific tasks,\nbetter meet desired performance characteristics, or adhere to\nsafety constraints.\n(19) Senior.--The term ``senior'' means an individual that\nhas attained 65 years of age.\n(20) Substantial modifier.--The term ``substantial\nmodifier'' means an entity that significantly alters (as\ndefined by the Commission) the behavior or capabilities of a\ncovered model through post training and inference enhancements.\n(21) Synthetic data.--The term ``synthetic data'' means\ndata that is artificially manufactured, typically by a computer\nprogram.\n(22) Vulnerable user.--The term ``vulnerable user'' means a\nminor, a senior, an individual with a disability, or an\nindividual experiencing a mental health crisis.\n\nSEC. 3. GUIDELINES FOR MODEL VERSIONING AND OTHER DESIGNATIONS.\n\n(a) Model Versioning.--Not later than 150 days after the date of\nenactment of this Act, the Commission, in consultation with the\nDirector of the National Institute of Standards and Technology, shall\nissue guidance regarding when a developer has produced a distinct\ngenerative artificial intelligence model through--\n(1) development, including novel pre-training of a model\narchitecture; or\n(2) any update to, or improvement or iteration of, an\nexisting generative artificial intelligence model.\n(b) Designation Process.--\n(1) In general.--Not later than 150 days after the date of\nenactment of this Act, the Commission, in consultation with the\nDirector of the National Institute of Standards and Technology,\nshall promulgate regulations in accordance with section 553 of\ntitle 5, United States Code, to establish criteria--\n(A) to determine when a deployment of a generative\nartificial intelligence model is expected to have\nsufficient users to meet the criteria described in--\n(i) paragraph (6) of section 2 for a\ndeployer to be considered a covered deployer;\nor\n(ii) paragraph (7) of section 2 for a\ngenerative artificial intelligence model to be\nconsidered a covered model; and\n(B) to designate a generative artificial\nintelligence model or a particular deployment of a\ngenerative artificial intelligence model as a covered\nmodel or covered system, based on a determination that\nit poses a significant risk to safety of minors,\nvulnerable users, or the public.\n(2) Requirements.--In establishing the criteria to make a\ndesignation under paragraph (1)(B), the Commission, in\nconsultation with the Director of the National Institute of\nStandards and Technology, shall do the following:\n(A) Consider the capabilities of the generative\nartificial intelligence model, the contexts in which\nthe model is deployed or reasonably likely to be\ndeployed (including the degree of access by vulnerable\nusers), and the severity and reversibility of the harms\nreasonably foreseeable from those capabilities.\n(B) Accompany any such designation with written\nfindings, which shall be made publicly available.\n(C) Provide the developer of the model or the\ndeployer of the system with written notice and not\nfewer than 30 days to respond before any such\ndesignation takes effect, except that a designation may\ntake effect immediately upon a finding by the\nCommission of imminent and significant risk, in which\ncase the Commission shall provide such notice and\nopportunity to respond not later than 30 days after the\neffective date and shall confirm, modify, or rescind\nthe designation not later than 60 days after the\neffective date.\n\nSEC. 4. ARTIFICIAL INTELLIGENCE MODEL DISCLOSURE MANDATES FOR\nDEVELOPERS.\n\n(a) Regulations.--\n(1) In general.--Not later than 150 days after the date of\nenactment of this Act, the Commission, in consultation with the\nDirector of the National Institute of Standards and Technology,\nshall promulgate regulations in accordance with section 553 of\ntitle 5, United States Code, requiring each developer of a\ncovered model (including a distinct generative artificial\nintelligence model that meets the definition of a covered\nmodel) to make publicly available the information described in\nsubsections (b) and (c)--\n(A) in a form determined by the Commission that\nincludes 2 separate disclosures, including--\n(i) 1 disclosure containing information\nrelevant to, and sufficiently detailed to\nfacilitate clear understanding for, a consumer;\nand\n(ii) 1 disclosure containing all\ninformation described in subsections (b) and\n(c) sufficiently detailed to permit meaningful\nreview by an independent researcher; and\n(B) in a manner that--\n(i) is accurate, materially complete, not\nmisleading by reason of a material omission,\nand adequate under standards determined by the\nCommission;\n(ii) does not rely primarily on generalized\nor conclusory language;\n(iii) facilitates transparency, public\ntrust, and independent evaluation;\n(iv) makes the information useful and\nactionable, including that the information is--\n(I) reliably available in a clear\nand conspicuous section of the website\nof the developer that can be easily\nlocated and accessed by a consumer in\nalignment with the Plain Writing Act of\n2010 (5 U.S.C. 301 note);\n(II) searchable; and\n(III) available through an\napplication programming interface;\n(v) does not disclose trade secrets; and\n(vi) does not expose classified systems or\ncreate national security risks.\n(2) Consultation.--In promulgating the regulations under\nparagraph (1), the Commission and the Director of the National\nInstitute of Standards and Technology shall engage relevant\nFederal agencies (including the Department of Health and Human\nServices, the Substance Abuse and Mental Health Services\nAdministration, and the National Institute of Mental Health)\nand third-parties with expertise relating to consumer\nprotection, the safety of minors, behavioral science, mental\nhealth, and human-computer interaction to ensure that the\ndisclosure required by such regulations reflects best practices\nand includes the appropriate information.\n(b) Initial Public Disclosure.--\n(1) In general.--The information described in this\nsubsection shall include the following:\n(A) Information relating to consumer transparency,\nsafety protocols for vulnerable users, and other\npractices associated with the covered model,\nincluding--\n(i) an overview of preventative safety\nsafeguards for minors implemented during pre-\ndeployment testing and model development;\n(ii) an overview of measures taken to\nprotect user privacy (including the privacy of\nminors) in training data, such as through data\nfiltering or anonymization methods, and\nmeasures taken to protect user privacy\n(including the privacy of minors) on covered\nplatforms;\n(iii) an overview of measures taken to\nensure the safety of minors in training data,\nsuch as through data filtering of child sexual\nabuse material;\n(iv) an overview of privacy standards\npertaining to the management, use, storage, and\npotential sale of user data (including the data\nof minors), including whether user data is used\nfor targeted advertising, user safety, internal\nor external research, and model training;\n(v) the durability and efficacy of\nvulnerable user safeguards (including\npreventing the generation of child sexual abuse\nmaterial and non-consensual intimate imagery),\nas well as the resiliency and scope of such\nfeatures;\n(vi) an overview of user and company\nreporting procedures for model behavior that\nposes a credible risk of material assistance to\nsuicidal ideation or suicide planning, or that\nposes a credible risk of material assistance to\nharm to others, including model behavior\nlearned in the post-deployment phase, and\nwhether such procedures were informed by crisis\nintervention experts and established standards;\nand\n(vii) as appropriate, the risk of negative\nimpacts for vulnerable users on topics\npertaining to mental health and well-being,\nincluding disclosure of research on the short-\nterm and long-term emotional and psychological\nimpacts of use for vulnerable users, and\nmethodological standards for such research.\n(B) System overviews of the covered model,\nincluding--\n(i) a high-level description of the\nintended, unintended, and prohibited uses of\nthe covered model, as reasonably anticipated by\nthe developer;\n(ii) a high-level description, including\nabilities and limitations, of the covered\nmodel;\n(iii) if applicable, a high-level overview\nof any difference between the covered model\nthat is the subject of the disclosure and most\nrecent previous version of the covered model\nthat is considered a distinct generative\nartificial intelligence model according to the\nguidance issued under section 3;\n(iv) a list of each covered platform and\ninterface through which the covered model is\naccessible;\n(v) a list of input and output formats\nsupported by the covered model; and\n(vi) an overview of the ability of the\ncovered model to trigger actions or\ncomputations external to the covered model\nduring model inference.\n(C) An overview of the post training and inference\nenhancements of the covered model and relevant\nmethodologies.\n(D) Information relating to safety, security, and\nevaluation protocols and practices associated with the\ncovered model, including--\n(i) a high-level overview of procedures or\nguiding principles for--\n(I) determining what constitutes\ndesirable model behavior;\n(II) ensuring desirable covered\nmodel behavior; and\n(III) preventing undesirable\ncovered model behavior;\n(ii) a description of methods or approaches\nused to evaluate and mitigate the risk that a\nmodel or the outputs of a related system are\nlower in quality or accuracy than intended use\ncases would require or expect;\n(iii) subject to subsection (d), an\noverview of measures taken to protect the\ncovered model and training data from attack or\nmisuse, including--\n(I) traditional cybersecurity\nprotocols, such as access controls and\nmulti-factor authentication; and\n(II) protections unique to AI\nsystems, such as preventing model\nweight exfiltration, data poisoning,\nand prompt injections;\n(iv) with respect to open weight models, an\noverview of the measures taken to protect\ntraining data from attack;\n(v) an overview of the criteria by which\nthe developer chose to publicly release the\ncovered model, such as evaluation against a\nrisk framework;\n(vi) a description of each of the\nevaluation criteria used to assess the risk\nprofile of the covered model, which shall\naddress, at a minimum, chemical, biological,\nradiological, and nuclear risk, offensive cyber\nrisk, autonomous loss-of-control risk, and any\nsimilar additional risk domain specified by the\nCommission, and which shall include, for any\nsuch domain in which the developer conducted no\nevaluation, an affirmative statement to that\neffect;\n(vii) an overview of the testing procedures\nand methods for achieving risk profile\nevaluations, such as AI red-teaming,\ncollaboration with external subject matter\nexperts, or solicitation of reports of model\nvulnerabilities from independent evaluators;\nand\n(viii) an overview of the performance of\nthe covered model on common public benchmarks\ndetermined by the Commission in coordination\nwith the Director of the National Institute of\nStandards and Technology, including any related\ninformation and in a form necessary for\nindependent researchers to verify the\nperformance of the covered model on each\nbenchmark, such as prompts and configuration\nsettings.\n(2) Timing of disclosure.--After the date on which the\nCommission issues guidance under section 3, a developer shall\nmake publicly available the information described in paragraph\n(1)--\n(A) with respect to the most recent version of a\ncovered model released to the public during the 2-year\nperiod before such date, not later than 90 days after\nsuch date;\n(B) with respect to a covered model released on or\nafter such date, concurrent with such release; or\n(C) with respect to the most recent version of a\ndistinct generative artificial intelligence model\nreleased on or after such date that, at a time after\nits release, meets the definition of a covered model,\nnot later than 90 days after such time.\n(c) Subsequent Public Disclosures.--\n(1) Requirement.--The information described in paragraph\n(2), with respect to a covered model, shall be made available\nto the public at such time, as determined by the Commission,\nand on an ongoing basis until the Commission determines that\nsuch disclosure is no longer necessary.\n(2) Information required.--The information described in\nthis paragraph is as follows:\n(A) Any significant change (as defined by the\nCommission in coordination with the Director of the\nNational Institute of Standards and Technology,\nincluding any substantial modification to, or observed\ndifferences from, the most recent prior public\ndisclosure under this section, and not including any\nimmaterial variation for which a disclosure requirement\nwould overly burden platform operations) to any of the\ninformation described in subsection (b) from the\ninformation reported in the initial information\ndisclosure, or, if applicable, any information reported\nin a subsequent disclosure.\n(B) Information including--\n(i) data dictionaries or equivalent\ndocumentation that explain significant\ndatasets, as defined by the Commission, in the\npossession of a developer related to the\ncovered model, including datasets that the\ndeveloper maintains relating to the users of\nthe relevant covered platforms, usage of the\ncovered platforms, content generated on covered\nplatforms, and actions taken to enforce the\nterms of service of covered platforms; and\n(ii) information relating to usage of the\ncovered model, including--\n(I) an overview of the most common\nviolations of covered model policies\nregarding acceptable use, including\ncommon violations pertaining to minor\nsafety;\n(II) an overview of how often the\nmodel refuses a prohibited prompt and\nthe developer has taken moderation\nactions in response to violations of\nacceptable use policies, including what\nthose actions are;\n(III) an overview of information\npertaining to attempts by users to\nutilize covered models to generate\nchild sexual abuse material or non-\nconsensual intimate imagery; and\n(IV) an overview of known system\ncapabilities for the utilization of the\ncovered model to develop or advance\ncyber, chemical, biological,\nradiological, and nuclear threats.\n(d) Special Rules for Disclosures.--\n(1) In general.--With respect to any disclosure required\nunder clause (iii) or (vi) of subsection (b)(1)(D) or under\nsubsection (c)(2)(B)(ii), such requirement shall not be\ninterpreted to require a developer to disclose any information\nthat, as determined by the Commission--\n(A) would compromise cybersecurity practices or\nprotections or safety practices or protections of the\ndeveloper; or\n(B) could provide material assistance to the\ndevelopment, acquisition, or deployment of chemical,\nbiological, radiological, nuclear, or offensive cyber\nweapons or capabilities.\n(2) Rule of construction.--Paragraph (1) shall not be\nconstrued to permit a developer to withhold--\n(A) the fact that an evaluation in a risk domain\ndescribed in subsection (b)(1)(D)(vi) was or was not\nconducted; or\n(B) a general characterization of the results of\nany such evaluation, at a level of detail that does not\ninclude information described in paragraph (1).\n(e) Reporting Requirement for Substantial Modifiers.--Not later\nthan 1 year after the date of enactment of this Act, the Commission, in\nconsultation with the Director of the National Institute of Standards\nand Technology, shall promulgate regulations in accordance with section\n553 of title 5, United States Code, requiring each substantial modifier\nto make publicly available the information described in subsections (b)\nand (c), as the Commission determines appropriate with respect to a\nsubstantial modifier.\n(f) Resources.--The Commission shall provide covered platforms with\ntechnical assistance to support compliance with the requirements of\nthis Act, particularly with respect to small and mid-sized platforms,\nas determined by the Commission.\n\nSEC. 5. MANDATORY DISCLOSURES FOR COVERED DEPLOYERS OF COVERED SYSTEMS.\n\nNot later than 150 days after the date of enactment of this Act,\nthe Commission, in consultation with the Director of the National\nInstitute of Standards and Technology, shall promulgate regulations in\naccordance with section 553 of title 5, United States Code, to\nestablish disclosure requirements for covered deployers with respect to\ncovered systems consistent with the requirements for developers of\ncovered models described in section 4, as determined by the Commission\nto be appropriate for a covered deployer with respect to a covered\nsystem.\n\nSEC. 6. ENFORCEMENT.\n\n(a) Unfair or Deceptive Acts or Practices.--A violation of section\n4 or 5 or a regulation promulgated under this Act shall be treated as a\nviolation of a rule defining an unfair or deceptive act or practice\nunder section 18(a)(1)(B) of the Federal Trade Commission Act (15\nU.S.C. 57a(a)(1)(B)).\n(b) Powers of the Commission.--\n(1) In general.--The Commission shall enforce this Act in\nthe same manner, by the same means, and with the same\njurisdiction, powers, and duties as though all applicable terms\nand provisions of the Federal Trade Commission Act (15 U.S.C.\n41 et seq.) were incorporated into and made a part of this Act.\n(2) Privileges and immunities.--Any person who violates\nsection 4 or 5 or a regulation promulgated under this Act shall\nbe subject to the penalties and entitled to the privileges and\nimmunities provided in the Federal Trade Commission Act (15\nU.S.C. 41 et seq.).\n(3) Authority preserved.--Nothing in this Act shall be\nconstrued to limit the authority of the Commission under any\nother provision of law.\n(4) Rulemaking.--Not later than 1 year after the date of\nenactment of this Act, the Commission shall promulgate in\naccordance with section 553 of title 5, United States Code,\nsuch rules as may be necessary to carry out this Act.\n\nSEC. 7. SEVERABILITY.\n\nIf any provision of this Act, or the application of such provision\nto any person or circumstance, is held to be unconstitutional, the\nremainder of this Act, and the application of the remaining provisions\nof this Act to any person or circumstance, shall not be affected.\n<all>","changes":[],"passport":{"data":{"act":{"jurisdiction":"США","title_official":"AI Systems Transparency Act","title_short":"Акт о прозрачности систем искусственного интеллекта","level":"законопроект","date_adopted":"2026-09-23 (внесение)","date_in_force":"","date_version":"","phased":"","status":"законопроект: на рассмотрении","sunset":"","regulator":"Федеральная торговая комиссия США","related":""},"goal":{"problem":"недостаточная прозрачность и безопасность систем искусственного интеллекта, особенно связанных с детьми и уязвимыми пользователями.","goal":"обеспечить прозрачность и безопасность систем искусственного интеллекта путем обязательного раскрытия информации и установления критериев безопасности.","targets":"отсутствуют","scope":"разработчики и операторы систем искусственного интеллекта, использующие открытые модели или имеющие значительное количество пользователей.","exclusions":"отсутствует"},"subjects_note":{"protected":"дети, уязвимые пользователи"},"subjects":[{"role":"поставщик, платформа, оператор, потребитель, госорган","who":"разработчики, операторы, платформы, потребители, госорганы","criteria":"оборот, численность пользователей, возраст пользователей","count":""}],"norms":[{"address":"ст. 6 ч. 1","addressee":"госорганы","essence":"Федеральная торговая комиссия вправе применять правила Федерального торгового кодекса к нарушениям данного акта.","type":"полномочие","mechanism":"административный контроль","cost_channel":"административные","cost_kind":"разовые","trigger":"обнаружение нарушений","sanction":"штраф за нарушение правил Федерального торгового кодекса","refs":"Федеральный торговый кодекс США","form":"нет установки","in_force":"","ru_analog":"[ПРОБЛЕМЫ]\nПроблемы: недостаток прозрачности и безопасности систем искусственного интеллекта, риски для детей и уязвимых групп населения"}]},"made_by":"GigaChat-2","made_at":"2026-10-07 09:13:02","edited_at":null,"edited_by":null}}