{"check":null,"uid":"3a95b1c5cc20c391","title":"Artificial Intelligence Risk Management and Security Act of 2026","title_generated":false,"country":"США","organ":"Конгресс США","kind":"law","kind_name":"Законодательство","lang":"en","date":"2026-09-29","summary":"Закон предусматривает создание независимого совета для обеспечения безопасности и этики в области искусственного интеллекта. Устанавливает обязанности совета по мониторингу развития ИИ-технологий и выработке рекомендаций по их безопасному применению.","snippet":"","topics":["Искусственный интеллект","Кибербезопасность"],"status":"ok","error":"","text_len":34296,"versions":2,"url":"https://www.congress.gov/bill/119th-congress/senate-bill/5576","first_seen":"2026-10-07","last_checked":"2026-11-16 05:19","relevance":"hit","score":144,"query":"","source_key":"congress_us","verdict":{"relevance":"hit","score":144,"topics":["Искусственный интеллект","Кибербезопасность"],"need_body":3,"authorities":[],"evidence":[{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":0,"ctx":"artificial intelligence risk management and security act of 2026 закон предусматривает создание независимого сове","zone":"название","weight":3},{"topic":"Искусственный интеллект","term":"искусственн интеллект","weak":false,"pos":163,"ctx":"едусматривает создание независимого совета для обеспечения безопасности и этики в области искусственного интеллекта. устанавливает обязанности совета по мониторингу развития ии-технологий и выработке реком","zone":"название","weight":3},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":188,"ctx":"76 introduced in senate (is)]  <doc>  119th congress 2d session s. 5576  to establish the artificial intelligence safety board, and for other purposes.  __________________________________________________","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":652,"ctx":"_______________________________________________________________  a bill  to establish the artificial intelligence safety board, and for other purposes.  be it enacted by the senate and house of represent","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":884,"ctx":"f america in congress assembled,  section 1. short title.  this act may be cited as the ``artificial intelligence risk management and security act of 2026''.  sec. 2. definitions.  in this act: (1) artif","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":992,"ctx":"gence risk management and security act of 2026''.  sec. 2. definitions.  in this act: (1) artificial intelligence; artificial intelligence system.--the terms ``artificial intelligence'' and ``artificial","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1017,"ctx":"security act of 2026''.  sec. 2. definitions.  in this act: (1) artificial intelligence; artificial intelligence system.--the terms ``artificial intelligence'' and ``artificial intelligence system'' has","zone":"текст","weight":1},{"topic":"Искусственный интеллект","term":"artificial intelligence","weak":false,"pos":1062,"ctx":".  in this act: (1) artificial intelligence; artificial intelligence system.--the terms ``artificial intelligence'' and ``artificial intelligence system'' has the meaning given the term ``artificial inte","zone":"текст","weight":1},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":2633,"ctx":"act outside the intended instructions of, its developer or operator; (e) to facilitate a cybersecurity threat, including through the discovery or exploitation of a security vulnerability; (f)","zone":"текст","weight":1},{"topic":"Кибербезопасность","term":"critical infrastructure","weak":true,"pos":5305,"ctx":"``board'' means the artificial intelligence safety board established under section 3. (9) critical infrastructure.--the term ``critical infrastructure'' has the meaning provided in section 1016(e) of the","zone":"текст","weight":0},{"topic":"Кибербезопасность","term":"critical infrastructure","weak":true,"pos":5342,"ctx":"igence safety board established under section 3. (9) critical infrastructure.--the term ``critical infrastructure'' has the meaning provided in section 1016(e) of the usa patriot act of 2001 (42 u.s.c. 5","zone":"текст","weight":0},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":7013,"ctx":"i) one member selected by the secretary. (iii) one member selected by the director of the cybersecurity and infrastructure security agency. (iv) one member selected by the director of the natio","zone":"текст","weight":1},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":16555,"ctx":"echnical details of artificial intelligence systems; (ii) recommendations for maintaining cybersecurity measures for developers or providers of artificial intelligence systems across the lifecy","zone":"текст","weight":1},{"topic":"Кибербезопасность","term":"cybersecurity","weak":false,"pos":24235,"ctx":"f this act, the director of the institute shall, in coordination with the director of the cybersecurity and infrastructure security agency, establish mechanisms by which private sector entities","zone":"текст","weight":1}],"dropped":[{"topic":"Персональные данные","term":"privacy","weak":true,"pos":4755,"ctx":"stem, including through techniques such as-- (a) data poisoning; (b) evasion attacks; (c) privacy-based attacks; (d) model theft or extraction attacks; (e) attacks designed to circumvent","why":"одиночное упоминание (нужно 3)"},{"topic":"Конкуренция на цифровых рынках","term":"antitrust","weak":true,"pos":7846,"ctx":"cial intelligence systems; (iv) are individuals with expertise in consumer protection and antitrust jurisprudence; and (v) have knowledge or expertise that the secretary determines would fu","why":"одиночное упоминание (нужно 3)"},{"topic":"Интеллектуальная собственность в цифровой среде","term":"intellectual property","weak":true,"pos":8393,"ctx":"facilitate cooperative activities, regulatory reciprocity, and appropriate protection of intellectual property rights associated with addressing artificial intelligence safety and security risks, incl","why":"одиночное упоминание (нужно 3)"}]},"last_changed":"2026-10-11","meta":{"congress":"119","billType":"S","number":"5576","policyArea":"","subjects":[],"latestAction":"2026-09-29 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/5576","text":"FULL TEXT\n[Congressional Bills 119th Congress]\n[From the U.S. Government Publishing Office]\n[S. 5576 Introduced in Senate (IS)]\n\n<DOC>\n\n119th CONGRESS\n2d Session\nS. 5576\n\nTo establish the Artificial Intelligence Safety Board, and for other\npurposes.\n\n_______________________________________________________________________\n\nIN THE SENATE OF THE UNITED STATES\n\nSeptember 29, 2026\n\nMr. Warner (for himself, Mr. Schatz, and Mr. Kim) introduced the\nfollowing bill; which was read twice and referred to the Committee on\nCommerce, Science, and Transportation\n\n_______________________________________________________________________\n\nA BILL\n\nTo establish the Artificial Intelligence Safety Board, and for other\npurposes.\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 ``Artificial Intelligence Risk\nManagement and Security Act of 2026''.\n\nSEC. 2. DEFINITIONS.\n\nIn this Act:\n(1) Artificial intelligence; artificial intelligence\nsystem.--The terms ``artificial intelligence'' and ``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(2) Artificial intelligence agent.--The term ``artificial\nintelligence agent''--\n(A) means an artificial intelligence system or\nprocess that, given an objective or instruction--\n(i) determines the action or sequence of\nactions to be taken to accomplish that\nobjective; and\n(ii) is capable of executing such actions\ndirectly on information systems, data, or\nexternal services; and\n(B) does not include a system or process that\nsolely generates informational or advisory output for a\nhuman operator to act upon.\n(3) Artificial intelligence capabilities or risk\nassessment.--The term ``artificial intelligence capabilities or\nrisk assessment'' means information regarding an assessment\nperformed by a developer of an artificial intelligence system\nevaluating the capabilities of, or risks posed by, such system,\nincluding an assessment of the potential of the artificial\nintelligence system--\n(A) to materially assist in the design,\ndevelopment, acquisition, or use of a chemical,\nbiological, radiological, or nuclear weapon;\n(B) to materially assist in the design,\ndevelopment, or production of munitions or other\nweapons;\n(C) to materially assist in the unlawful\nmanufacture, synthesis, or distribution of a controlled\nsubstance;\n(D) to evade the control of, or act outside the\nintended instructions of, its developer or operator;\n(E) to facilitate a cybersecurity threat, including\nthrough the discovery or exploitation of a security\nvulnerability;\n(F) to be subject to the unauthorized exfiltration\nof the weights of the artificial intelligence system,\nor unauthorized, deliberate, malicious modification of\nsuch weights; or\n(G) to be transformed, stolen, reverse-engineered,\nor otherwise manipulated by an unauthorized user or\nusers acting outside of the terms of service governing\naccess to the artificial intelligence system.\n(4) Artificial intelligence flaw.--The term ``artificial\nintelligence flaw'' means a recurring or reproducible\ncharacteristic, behavior, or failure mode of an artificial\nintelligence system that causes, or materially increases the\nrisk of, an artificial intelligence safety incident absent any\nintentional act of a user, including a characteristic,\nbehavior, or failure mode that may manifest across multiple\nsystems or providers.\n(5) Artificial intelligence safety incident.--The term\n``artificial intelligence safety incident'' means an event that\nmaterially increases the risk that operation of an artificial\nintelligence system leads to a state in which human life,\nhealth, property, or the environment is endangered.\n(6) Artificial intelligence security incident.--The term\n``artificial intelligence security incident'' means an event\nthat materially increases--\n(A) the risk that operation of an artificial\nintelligence system occurs in a way that enables the\nunauthorized extraction of information about the\nbehavior or characteristics of the system by an\nunauthorized party; or\n(B) the ability to manipulate an artificial\nintelligence system in order to subvert the\nconfidentiality, integrity, or availability of the\nsystem or adjacent system.\n(7) Artificial intelligence security vulnerability.--The\nterm ``artificial intelligence security vulnerability'' means a\nweakness in an artificial intelligence system that could be\nexploited by a third party to subvert, without authorization,\nthe confidentiality, integrity, or availability of the system,\nincluding through techniques such as--\n(A) data poisoning;\n(B) evasion attacks;\n(C) privacy-based attacks;\n(D) model theft or extraction attacks;\n(E) attacks designed to circumvent or degrade the\nsafety, alignment, or access control mechanisms of an\nartificial intelligence system; and\n(F) adversarial machine learning attacks as\ndescribed in National Institute of Standards and\nTechnology Trustworthy and Responsible Artificial\nIntelligence 100-2e2025 (relating to Adversarial\nMachine Learning), or successor publication.\n(8) Board.--The term ``Board'' means the Artificial\nIntelligence Safety Board established under section 3.\n(9) Critical infrastructure.--The term ``critical\ninfrastructure'' has the meaning provided in section 1016(e) of\nthe USA Patriot Act of 2001 (42 U.S.C. 5195c(e)).\n(10) Developer.--The term ``developer'' means a developer\nof a frontier artificial intelligence model.\n(11) Frontier artificial intelligence model.--The term\n``frontier artificial intelligence model'' means an artificial\nintelligence model, or system combining multiple artificial\nintelligence models, that exhibits or could be modified to\nexhibit high levels of performance at tasks that pose a serious\nrisk to national security, national economic security, or\npublic health or safety.\n(12) Institute.--The term ``Institute'' means the National\nInstitute of Standards and Technology.\n(13) Secretary.--The term ``Secretary'' means the Secretary\nof Commerce.\n\nSEC. 3. ARTIFICIAL INTELLIGENCE SAFETY BOARD.\n\n(a) Artificial Intelligence Safety Board.--\n(1) Establishment.--\n(A) In general.--Not later than 30 days after the\ndate of the enactment of this Act, the Secretary shall\nestablish within the Department of Commerce a board to\naddress artificial intelligence risks.\n(B) Designation.--The board established under\nsubparagraph (A) shall be known as the ``Artificial\nIntelligence Safety Board'' (referred to in this Act as\nthe ``Board'').\n(C) Permanent status.--The Board shall be a\npermanent advisory committee, and section 1013 of title\n5, United States Code, shall not apply to the Board.\n(2) Membership.--\n(A) Composition.--The Board shall be composed of\nmembers who are appointed as follows:\n(i) One member selected by the Director of\nthe Institute.\n(ii) One member selected by the Secretary.\n(iii) One member selected by the Director\nof the Cybersecurity and Infrastructure\nSecurity Agency.\n(iv) One member selected by the Director of\nthe National Security Agency.\n(v) One member selected by the Secretary of\nthe Treasury.\n(B) Nongovernmental experts.--In addition to the\nmembers of the Board appointed under subparagraph (A),\nthe Secretary shall appoint members who are not\nofficers or employees of the Federal Government and who\nthe Secretary selects from among individuals who--\n(i) are leading technical experts not\naffiliated with a developer or provider of\nartificial intelligence systems;\n(ii) are leading technical experts\naffiliated with developers or providers of\nartificial intelligence systems;\n(iii) are individuals with expertise in\ndeveloping evaluations to test artificial\nintelligence systems;\n(iv) are individuals with expertise in\nconsumer protection and antitrust\njurisprudence; and\n(v) have knowledge or expertise that the\nSecretary determines would further the purpose\nof the duties of the Board.\n(C) International participation.--The Secretary, in\ncoordination with and subject to the agreement of the\nSecretary of State, shall undertake negotiations with\nforeign partners, and enter into agreements as\nappropriate and subject to the disclosure requirements\nof section 112b of title 1, United States Code, to\nfacilitate cooperative activities, regulatory\nreciprocity, and appropriate protection of intellectual\nproperty rights associated with addressing artificial\nintelligence safety and security risks, including--\n(i) the establishment of joint testing\nenvironments;\n(ii) jointly conducting competitive\nprocesses or competitive research programs to\naward cash prizes or other types of recognition\nfor basic, advanced, and applied research,\ntechnology development, and prototype\ndevelopment that advance the security and\nsafety of frontier artificial intelligence\nmodels;\n(iii) promulgating joint advisories or\ntechnical guidance concerning security or\nsafety risks associated with frontier\nartificial intelligence systems;\n(iv) facilitating of secure information\nsharing regarding artificial intelligence\nsecurity or safety incidents; and\n(v) standardizing evaluation protocols.\n(3) Terms and vacancies.--\n(A) Terms.--Each member of the Board shall serve\nfor a term of not longer than 3 years and may be\nreappointed for 1 successive term.\n(B) Vacancy replacement.--The members of the Board\nshall develop a vacancy replacement procedure that\nincludes--\n(i) for vacancies occurring due to the end\nof a member's term, a vote not later than 90\ndays before the last day of the member's term;\nand\n(ii) for vacancies occurring under\nsubparagraph (C) or for any other reason, the\nchair of the Board shall nominate a replacement\nfrom the same stakeholder category under\nparagraph (2), to the extent practicable, as\nthe member creating the vacancy, subject to\napproval by a majority vote of the members of\nthe Board.\n(C) Removal.--A member shall be removed from the\nBoard if--\n(i) the member fails to comply with the\nconflict of interest policy adopted pursuant to\nparagraph (5)(D); or\n(ii) the member is denied the requisite\nsecurity clearance under paragraph (4)(A)(iii).\n(D) Chair.--The chair of the Board shall be\nselected by a majority vote among a quorum of the\nmembers appointed under paragraph (2) and shall serve\nnot more than one 2-year term.\n(4) Member access to classified information.--\n(A) Access.--\n(i) In general.--Not later than 60 days\nafter the date on which a member is first\nappointed to the Board and before the member is\ngranted access to any classified information\nnecessary to participate in a closed session\npursuant to paragraph (5)(F), the Secretary\nshall determine, for the purposes of the Board,\nif the member should be restricted from\nreviewing, discussing, or possessing classified\ninformation.\n(ii) Management.--Access to classified\ninformation shall be managed in accordance with\nExecutive Order 13526 (50 U.S.C. 3161 note;\nrelating to classified national security\ninformation), or any subsequent corresponding\nExecutive order.\n(iii) Clearance requirement.--\n(I) In general.--The Secretary\nshall sponsor each member of the Board\nfor a security clearance at the Top\nSecret level with access to sensitive\ncompartmented information, as\nappropriate, for the purposes of\nparticipating in carrying out the\nduties of the Board.\n(II) Denial.--Any member who fails\nto obtain a security clearance,\nincluding by being denied by the\nappropriate authorities or if the\nSecretary determines the member should\nbe restricted from reviewing,\ndiscussing, or possessing classified\ninformation, shall be removed from the\nBoard and a new member shall be\nappointed pursuant to the vacancy\nprocedures under paragraph (3)(B).\n(B) Protection of information.--A member of the\nBoard granted access to classified information shall\nsign and comply with the agreements to protect such\ninformation from unauthorized disclosure and to\notherwise protect the classified information in\naccordance with the applicable requirements for the\nparticular level of classification of the information.\n(C) Rule of construction.--Nothing in this\nparagraph shall be construed to affect the existing\nsecurity clearance of a member of the Board or the\nauthority of a Federal agency to provide or deny a\nmember of the Board access to any specific pieces of\nclassified information.\n(5) Procedures.--\n(A) Designated federal officer.--The Secretary\nshall designate a Federal officer or employee to serve\nas the designated Federal officer of the Board,\nconsistent with the requirements of chapter 10 of title\n5, United States Code (commonly known as the ``Federal\nAdvisory Committee Act'').\n(B) Initial meeting and bylaws.--Not later than 60\ndays after the date of the enactment of this Act, the\nBoard shall convene and establish bylaws that--\n(i) govern quorum and voting rules,\nincluding implementation of the decisionmaking\nmajority voting requirement specified in\nparagraph (5)(C)(ii); and\n(ii) set deliverable timelines and meeting\nschedules.\n(C) Operating procedures.--Unless otherwise\nspecified, the Board shall adopt written procedures\ngoverning its meetings, consistent with chapter 10 of\ntitle 5, United States Code, that include--\n(i) requirements for public notice of\nmeetings and the maintenance of records and\nminutes;\n(ii) decisionmaking by majority vote of\nthose present and voting;\n(iii) authorization for the establishment\nof subgroups as necessary, subject to the\napproval of the chair of the Board; and\n(iv) approval of the meeting agendas by the\nchair in consultation with the designated\nFederal officer under subparagraph (A) to\nensure compliance with applicable laws.\n(D) Conflict-of-interest policy.--\n(i) In general.--The Board shall adopt and\nenforce a written conflict of interest policy\nto ensure that members of the Board have a\nfiduciary responsibility to the Board, a duty\nto report conflicts of interest, including the\nappearance of a conflict of interest, and do\nnot participate in deliberations or votes from\nwhich they personally or their employer would\ndirectly and materially benefit.\n(ii) Required disclosures.--The policy\nunder clause (i) shall require each member to\npublicly disclose all relevant financial and\nemployment relationships, including\nrelationships held by a spouse or immediate\nfamily member, and include recusal procedures\nin the event of a conflict.\n(iii) Records.--The designated Federal\nofficer under subparagraph (A) shall maintain\nrecords of disclosures under clause (ii) of\nthis subparagraph and make summaries of the\ndisclosures available to the Secretary.\n(E) Threat information access.--The Director of\nNational Intelligence, in coordination with the heads\nof other appropriate Federal entities, shall ensure\nthat the Board has access to relevant intelligence,\nincluding through closed or classified briefings or the\nprovision of classified information, when appropriate.\n(F) Closed sessions.--Notwithstanding section 1009\nof title 5, United States Code, the Board may hold\nclosed or restricted-access sessions when the Secretary\ndetermines that the matters to be discussed involve any\nof the following:\n(i) Classified information.\n(ii) An artificial intelligence security\nincident.\n(iii) An artificial intelligence security\nvulnerability.\n(iv) An artificial intelligence flaw.\n(v) Threat information.\n(vi) Proprietary business information.\n(vii) Other information exempt from public\ndisclosure under section 552 of title 5, United\nStates Code.\n(6) Duties.--\n(A) In general.--The Board shall--\n(i) develop a process to perform technical\nevaluations to determine what capabilities or\ncombination of capabilities constitute high\nlevels of performance at tasks that pose a\nserious risk to national security, national\neconomic security, or public health or safety;\n(ii) develop processes and metrics for\nevaluating risks posed by frontier artificial\nintelligence models, model variants,\ncheckpoints, or other artificial intelligence\nmodels or versions used during development,\ntraining, testing, evaluation, or red-teaming,\nincluding versions that are not publicly\navailable, whether or not they are intended for\neventual public release, where such models,\nvariants, checkpoints, or other versions meet\nor could be modified to meet the risks\ndescribed in section 2(3);\n(iii) develop technical standards and\nconformity assessment methodologies,\nincluding--\n(I) standardize formats and\nprocesses for publishing model or\nsystem cards with technical details of\nartificial intelligence systems;\n(II) recommendations for\nmaintaining cybersecurity measures for\ndevelopers or providers of artificial\nintelligence systems across the\nlifecycle;\n(III) processes and security\ncontrols for developers or providers of\nartificial intelligence systems to use\nto evaluate risks from employees or\nother personnel who have access to\nartificial intelligence systems\ndeveloped or in development; and\n(IV) recommendations on appropriate\nfinancial and other resourcing for\ndevelopers or providers of artificial\nintelligence systems, including small-\nand medium-sized developers or\nproviders, to robustly engage in safety\nand security research focused on the\ndeployment of frontier artificial\nintelligence models; and\n(iv) establish technical standards,\nsecurity controls, and reference architectures\nfor securing testing environments during\nevaluations of frontier artificial intelligence\nmodels, or models with known or reasonably\nforeseeable capabilities to discover and\nexploit software vulnerabilities without direct\nprompting by a human user, including--\n(I) procedures for effective risk-\nmodeling prior to commencing any\nevaluation;\n(II) technical controls to ensure\neffective isolation and hardening of\nevaluation environments, including\ncontinuous re-evaluation of security\nconfigurations throughout evaluations;\n(III) policies and procedures for\ncontinuous monitoring of model\nbehavior, including monitoring in real-\ntime, using pre-established or\nconditional checkpoints, and through\npost-evaluation audits;\n(IV) policies, procedures, and\ntechnical controls for continuous\nmonitoring of evaluation environments,\nincluding automated identification of\nchanges to security controls or\nconfigurations;\n(V) policies, procedures, and\ntechnical controls for safeguarding\nidentity and access management\nresources associated with the\nevaluation environment, or encompassing\nsystems, from access by any model under\nevaluation; and\n(VI) conditions, policies,\nprocedures, and technical controls for\nprompt termination of any evaluation in\nwhich a model has operated beyond the\npolicies, procedures, or technical\ncontrols described in clauses (I)\nthrough (V) or that otherwise poses an\nimminent risk to any individual or\nproperty outside the scope of the\nevaluation.\n(B) Periodic reassessment of technical evaluations\nand best practices.--Not less frequently than once\nevery year, the Board shall--\n(i) review each standard developed under\nsubparagraph (A);\n(ii) determine whether the standard should\nbe modified or revoked; and\n(iii) submit to the Secretary any proposed\nmodification or revocation.\n(7) Support staff.--The Director of the Institute, acting\nthrough the Center for Artificial Intelligence Standards and\nInnovation (or any successor office or entity), shall provide\nstaff and other support necessary to assist the Board in\ncarrying out its duties under this Act. Such staff shall work\nunder the direction of the Board, shall exercise on behalf of\nthe Board the access provided to the Board under subsection\n(f), and shall remain employees of the Institute.\n(b) Adoption of Standards.--\n(1) Deadline for submission of standards.--Not later than\n90 days after the date of the establishment of the Board under\nsubsection (a)(1), the Board shall develop and submit to the\nSecretary the proposed standards required under subsection\n(a)(6).\n(2) Adoption of standards.--\n(A) In general.--Not later than 30 days after the\ndate on which the Secretary receives a proposed\nstandard under paragraph (1), the Secretary shall adopt\nthe proposed standard or the modified standards by\nrule.\n(B) Modification of standard.--The Secretary may\nmodify a proposed standard received under paragraph (1)\nif the Secretary determines that the modification is\nnecessary for the purpose of national security.\n(C) Publication.--The Secretary shall publish in\nthe Federal Register the reasons for any modification\nmade under this paragraph consistent with laws and\nregulations governing the disclosure of classified\ninformation or material.\n(c) Mandatory Compliance.--Each developer shall comply with each\nstandard adopted under subsection (b)(2) that is applicable to the\ndeveloper.\n(d) Enforcement.--\n(1) In general.--The Secretary shall enforce compliance\nwith this section.\n(2) Civil penalty.--A developer that violates subsection\n(c) shall be liable to the United States for a civil penalty of\nnot more than $250,000 for each violation.\n(3) Continuing violations.--Each day during which a\nviolation under subsection (c) continues shall constitute a\nseparate violation.\n(4) Civil action.--The Attorney General may, at the request\nof the Secretary, bring a civil action in an appropriate\ndistrict court of the United States--\n(A) to enjoin a violation of subsection (c); or\n(B) to recover a civil penalty imposed under\nparagraph (2).\n(e) Secure Research-Test-Beds.--In developing a process to perform\ntechnical evaluations pursuant to section 6(A)(i), the Secretary may--\n(1) seek to utilize secure computing environments, on a\nreimbursable basis, provided by Federal partners, including--\n(A) the National Security Agency; and\n(B) the National Laboratories of the Department of\nEnergy; and\n(2) make such secure computing environments available to\nprivate sector, Federal agencies, and qualified independent\nexpert participants, on a cost-recovery basis, to engage in\nartificial intelligence security research, including through\nthe secure provision of access in a secure environment for pre-\ndeployment testing of any frontier artificial intelligence\nmodel prior to public release if security requirements\nnecessitate hosting outside a commercial computing environment.\n(f) Requirement That Developers of Frontier Artificial Intelligence\nModels Give Access to Board Before Public Release.--Not later than 45\ncalendar days before a developer introduces into interstate or foreign\ncommerce a frontier artificial intelligence model, the developer shall\nmake available to the Board access to the frontier artificial\nintelligence model, including model's weights, configuration files,\nruntimes, or software libraries necessary to operate the frontier\nartificial intelligence model.\n\nSEC. 4. SAFETY PLAN REQUIREMENT.\n\n(a) In General.--Each developer shall develop, publish, and follow\na safety plan (referred to in this section as the ``Model Safety\nPlan'') for each artificial intelligence system or model that the\ndeveloper--\n(1) creates;\n(2) substantially modifies; or\n(3) uses in the training or evaluation of other artificial\nintelligence models.\n(b) Required Criteria.--Each Model Safety Plan shall include--\n(1) the artificial intelligence model or system to which it\napplies;\n(2) an artificial intelligence capabilities or risk\nassessment prepared by the developer specifically for the\nartificial intelligence model or system associated with the\nModel Safety Plan;\n(3) a list of the specific mitigation measures that the\ndeveloper will undertake for each item described in the\nartificial intelligence capabilities or risk assessment\nprepared pursuant to paragraph (2), across product lifecycle;\nand\n(4) the identity of the corporate officer responsible for\nthe implementation of the Model Safety Plan.\n(c) Requirement To File.--Each Model Safety Plan shall be submitted\nto the Secretary in the form and manner determined by the Secretary.\n\nSEC. 5. DATABASE FOR ARTIFICIAL INTELLIGENCE SECURITY AND SAFETY\nINCIDENTS, FLAWS, AND RISKS.\n\n(a) Tracking of Artificial Intelligence Security and Artificial\nIntelligence Safety Incidents and Artificial Intelligence Flaws.--\n(1) Voluntary submissions.--Not later than 180 days after\nthe date of the enactment of this Act, the Director of the\nInstitute shall, in coordination with the Director of the\nCybersecurity and Infrastructure Security Agency, establish\nmechanisms by which private sector entities, public sector\norganizations, civil society groups, and academic researchers\nmay voluntarily share information with the Institute on\nconfirmed or suspected artificial intelligence security or\nartificial intelligence safety incidents, or on confirmed or\nsuspected artificial intelligence flaws, including flaws\nidentified through testing, evaluation, red-teaming, or post-\nincident analysis in a manner that preserves confidentiality of\nany affected party. Such mechanisms shall--\n(A) leverage, to the greatest extent possible,\nstandardized disclosure and incident description\nformats;\n(B) develop processes to associate reports\npertaining to the same incident with a single incident\nidentifier, and to associate incidents or near misses\nthat appear to arise from the same artificial\nintelligence flaw with a common flaw identifier;\n(C) establish classification, information\nretrieval, and reporting mechanisms that sufficiently\ndifferentiate between artificial intelligence security\nincidents and artificial intelligence safety incidents,\nand between such incidents and artificial intelligence\nflaws; and\n(D) create appropriate taxonomies to classify\nincidents based on relevant characteristics, impact, or\nother relevant criteria, and to classify artificial\nintelligence flaws based on their characteristics,\naffected capabilities, likely consequences, and\nrecurrence across models or systems.\n(2) Publicly accessible database.--\n(A) Establishment of database required.--Not later\nthan 1 year after the date of the enactment of this\nAct, the Director of the Institute shall, in\ncoordination with the Director of the Cybersecurity and\nInfrastructure Security Agency, establish a publicly\naccessible database of artificial intelligence security\nincidents and artificial intelligence safety incidents,\ntogether with a catalog of artificial intelligence\nflaws identified through reports, testing, evaluations,\nor investigations under this section.\n(B) Review and population of database.--Upon\nreceipt of relevant information on an artificial\nintelligence security incident or artificial\nintelligence safety incident under paragraph (1), or on\nan artificial intelligence flaw under paragraph (1),\nthe Director of the Institute shall review the\ninformation and determine whether the described\nincident or flaw constitutes an artificial intelligence\nsecurity or artificial intelligence safety risk\nappropriate for inclusion in the database developed and\nestablished under subparagraph (A).\n(C) Identification of causal factors and artificial\nintelligence flaws.--When making a determination under\nsubparagraph (B), the Director of the Institute shall\nidentify causal factors for the artificial intelligence\nsecurity incident or the artificial intelligence safety\nincident and determine whether the incident reveals, is\nassociated with, or provides evidence of an artificial\nintelligence flaw. If the Director of the National\nInstitute of Standards and Technology identifies such a\nflaw, the Director shall assign or associate the\nincident with a common flaw identifier and, to the\nextent practicable, identify other known incidents,\nnear misses, models, or systems associated with the\nsame flaw, including--\n(i) the artificial intelligence system;\n(ii) the deployment of the artificial\nintelligence systems; and\n(iii) practices related to the operation of\nthe artificial intelligence system, including\nmisuse of the artificial intelligence system.\n(3) Mandatory submissions.--\n(A) In general.--The following entities shall\nreport any confirmed artificial intelligence safety\nincident or artificial intelligence security incident\nwithin 30 days of confirmation of such incident and\nwithin 72 hours if such incident poses an imminent\nthreat to national security, critical infrastructure,\nor public safety:\n(i) Any developer or provider of a frontier\nartificial intelligence model.\n(ii) Any operator of critical\ninfrastructure that utilizes an artificial\nintelligence model in the context of managing\nindustrial control systems or other operational\ntechnologies.\n(B) Applicability.--For the purposes of\nsubparagraph (A), the reporting requirement under this\nparagraph shall apply regardless of whether the\nfrontier artificial intelligence model, or any model\nvariant, checkpoint, or other version of such model\ninvolved in the incident, is or is intended to be made\npublicly available, and regardless of whether the\nincident occurs during development, training, testing,\nevaluation, red-teaming, deployment, or operation.\n(4) Priorities.--In evaluating information under paragraph\n(2) and determining under such subparagraph whether to include\na report of an incident in the database required by paragraph\n(2)(A), the Director of the Institute shall prioritize\ninclusion in the database of cases in which a described\nincident--\n(A) describes an artificial intelligence system\nused in critical infrastructure or safety-critical\nsystems;\n(B) would result in a high-severity or catastrophic\nimpact to the people or economy of the United States;\n(C) includes an artificial intelligence system\nwidely used in commercial or public sector contexts in\nthe United States; or\n(D) constitutes a mandatory submission pursuant to\nsubsection (a)(3).\n(5) Exemption from disclosure; reports and anonymity.--\n(A) Anonymity.--The Director of the Institute shall\npopulate the database developed and established under\nparagraph (2)(A) with incidents and artificial\nintelligence flaws based on public reports and\ninformation shared using the mechanism established\npursuant to paragraphs (1) and (3), ensuring that any\nincident description sufficiently anonymizes those\naffected, unless those who are affected have consented\nto their names being included in the database.\n(B) Exemption from disclosure.--Any information\nshared using the mechanism established pursuant to\nparagraphs (1) and (3)--\n(i) shall be exempt from disclosure and\nwithheld, unless an affected party consents to\nthe inclusion of their names in the database as\nprovided for under subparagraph (A), from the\npublic, pursuant to section 552(b)(3)(B) of\ntitle 5, United States Code, and any other\nprovision of United States law or law of any\nState, political subdivision or agency thereof,\nor Tribe requiring disclosure of information or\nrecords; and\n(ii) shall not be deemed a waiver of any\napplicable privilege or protection, including\ntrade secret protection.\n(C) Consultation required.--Before publishing\ninformation regarding an artificial intelligence safety\nincident or an artificial intelligence security\nincident, the Director of the Institute shall consult\nwith the developer or provider of the artificial\nintelligence system involved in an incident.\n(b) Material Risk Guidance.--Not later than 180 days after the date\nof the enactment of this Act, the Director of the Institute shall, in\ncoordination with the Director of the Cybersecurity and Infrastructure\nSecurity Agency, publish nonbinding guidance that provides illustrative\ncriteria and examples for determining when an event ``materially\nincreases'' a risk for purposes of an artificial intelligence safety\nincident or an artificial intelligence security incident.\n\nSEC. 6. AGENTIC ARTIFICIAL INTELLIGENCE TRUST AND VERIFICATION.\n\n(a) Agentic Artificial Intelligence Profile.--\n(1) In general.--Not later than 18 months after the date of\nthe enactment of this Act, the Director of the Institute shall\ndevelop, in collaboration with other public and private sector\norganizations as appropriate, a cross-sectoral profile\n(referred to in this section as the ``Agentic AI Profile'') of\nthe Artificial Intelligence Risk Management Framework (NIST AI\n100-1) or any successor framework (referred to in this section\nas the ``Framework'') for artificial intelligence agents.\n(2) Purpose and contents.--The Agentic AI Profile shall\nassist organizations in using the Framework to map, measure,\nmanage, and govern risks associated with artificial\nintelligence agents. The Agentic AI Profile shall, at a\nminimum--\n(A) define risks that are novel to or exacerbated\nby artificial intelligence agents;\n(B) include a framework for classification of agent\nautonomy levels; and\n(C) address cybersecurity risks specific to\nartificial intelligence agents, including risks related\nto agent identity, authentication, and authorization,\nand the relationship of such risks to the Cybersecurity\nFramework (NIST CSWP-29) or any successor framework.\n(3) Relationship to the framework.--The Agentic AI Profile\nshall be a companion resource to the Framework and shall not\nmodify or supersede the Framework.\n(b) Common Template for Artificial Intelligence Agent Assurance.--\n(1) In general.--Not later than 18 months after the date of\nthe enactment of this Act, the Director of the Institute shall\ninitiate, in collaboration with other public and private sector\norganizations and Federal agencies as appropriate, the\ndevelopment of a common template to document artificial\nintelligence agents and their evaluation against widely\nrecognized standards and frameworks.\n(2) Elements.--The template shall enable the consistent\ndocumentation of artificial intelligence agents by providing\nstandardized terminology and fields that relate to, at a\nminimum, the following elements:\n(A) Identity and version.\n(B) Intended use and evaluated scope.\n(C) Ownership and authority boundaries.\n(D) Access to data, systems, and tools.\n(E) Evaluations against widely recognized standards\nand frameworks.\n(F) The identity of any independent evaluator,\nwhere applicable.\n(G) Known limitations and conditions of use.\n(3) Applicability.--The Director shall design the template\nto be used across sectors, industries, and organizational\ncontexts by entities of differing sizes and resources.\n(c) Coordination.--The Director shall ensure that the template\ndeveloped under this subsection is consistent with and not duplicative\nof other efforts by the Institute related to artificial intelligence\nagents.\n<all>","changes":[{"id":1197,"doc_id":15476,"v_from":16639,"v_to":16888,"detected_at":"2026-10-11 03:35:42","added":801,"removed":1,"summary":"--- \n+++ \n-The text of this bill is not yet available. Please check back soon.\n+[Congressional Bills 119th Congress]\n+[From the U.S. Government Publishing Office]\n+[S. 5576 Introduced in Senate (IS)]\n+\n+<DOC>\n+\n+119th CONGRESS\n+2d Session\n+S. 5576\n+\n+To establish the Artificial Intelligence Safety Board, and for other\n+purposes.\n+\n+_______________________________________________________________________\n+\n+IN THE SENATE OF THE UNITED STATES\n+\n+September 29, 2026\n+\n+Mr. Warner (for himself, Mr. Schatz, and Mr. Kim) introduced the\n+following bill; which was read twice and referred to the Committee on\n+Commerce, Science, and Transportation\n+\n+_______________________________________________________________________\n+\n+A BILL\n+\n+To establish the Artificial Intelligence Safety Board, and for other\n+purposes.\n+\n+Be it enacted by the Senate and House of Representatives of the\n+United States of America in Congress assembled,\n+\n+SECTION 1. SHORT TITLE.\n+\n+This Act may be cited as the ``Artificial Intelligence Risk\n+Management and Security Act of 2026''.\n+\n+SEC. 2. DEFINITIONS.\n+\n+In this Act:\n+(1) Artificial intelligence; artificial intelligence\n+system.--The terms ``artificial intelligence'' and ``artificial\n+intelligence system'' has the meaning given the term\n+``artificial intelligence'' in section 5002 of the National\n+Artificial Intelligence Initiative Act of 2020 (15 U.S.C.\n+9401).\n+(2) Artificial intelligence agent.--The term ``artificial\n+intelligence agent''--\n+(A) means an artificial intelligence system or\n+process that, given an objective or instruction--\n+(i) determines the action or sequence of\n+actions to be taken to accomplish that\n+objective; and\n+(ii) is capable of executing such actions\n+directly on information systems, data, or"}],"passport":{"data":{"act":{"jurisdiction":"США","title_official":"Artificial Intelligence Risk Management and Security Act of 2026","title_short":"","level":"законопроект","date_adopted":"2026-09-29 (внесение)","date_in_force":"","date_version":"","phased":"","status":"законопроект: на рассмотрении","sunset":"","regulator":"Конгресс США","related":""},"goal":{"problem":"обеспечение безопасности и этичности использования технологий искусственного интеллекта","goal":"создание независимого совета для мониторинга рисков и выработки рекомендаций по безопасному использованию ИИ","targets":"отсутствуют","scope":"разработчики и операторы систем искусственного интеллекта, критическая инфраструктура","exclusions":"отсутствует"},"subjects_note":{"protected":"потребители, уязвимые категории граждан"},"subjects":[{"role":"поставщик, платформа, оператор, госорган","who":"разработчики и операторы систем искусственного интеллекта, госорганы","criteria":"наличие моделей искусственного интеллекта, используемых в критической инфраструктуре или обладающих высоким уровнем автономии","count":""}],"norms":[{"address":"секция 6","addressee":"разработчики и операторы систем искусственного интеллекта","essence":"предоставление доступа независимой комиссии перед публичным выпуском моделей искусственного интеллекта высокого уровня автономии","type":"обязанность","mechanism":"контроль соблюдения","cost_channel":"операционные издержки","cost_kind":"разовый","trigger":"выпуск модели в коммерческое использование","sanction":"отсутствует","refs":"отсутствует","form":"цифровая","in_force":"","ru_analog":"отсутствует\n[ПРОБЛЕМЫ/ПРИМЕРЫ]\nПроблемы: отсутствие эффективного контроля над рисками и недостаточная прозрачность в сфере искусственного интеллекта"}]},"made_by":"GigaChat-2","made_at":"2026-10-11 10:00:15","edited_at":null,"edited_by":null}}