Database for artificial intelligence security and safety incidents, flaws, and risks

(a) Tracking of artificial intelligence security and artificial intelligence safety incidents and artificial intelligence flaws (1) Voluntary submissions Not later than 180 days after the date of the enactment of 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, public sector organizations, civil society groups, and academic researchers may voluntarily share information with the Institute on confirmed or suspected artificial intelligence security or artificial intelligence safety incidents, or on confirmed or suspected artificial intelligence flaws, including flaws identified through testing, evaluation, red-teaming, or post-incident analysis in a manner that preserves confidentiality of any affected party. Such mechanisms shall— (A) leverage, to the greatest extent possible, standardized disclosure and incident description formats; (B) develop processes to associate reports pertaining to the same incident with a single incident identifier, and to associate incidents or near misses that appear to arise from the same artificial intelligence flaw with a common flaw identifier; (C) establish classification, information retrieval, and reporting mechanisms that sufficiently differentiate between artificial intelligence security incidents and artificial intelligence safety incidents, and between such incidents and artificial intelligence flaws; and (D) create appropriate taxonomies to classify incidents based on relevant characteristics, impact, or other relevant criteria, and to classify artificial intelligence flaws based on their characteristics, affected capabilities, likely consequences, and recurrence across models or systems. (2) Publicly accessible database (A) Establishment of database required Not later than 1 year after the date of the enactment of this Act, the Director of the Institute shall, in coordination with the Director of the Cybersecurity and Infrastructure Security Agency, establish a publicly accessible database of artificial intelligence security incidents and artificial intelligence safety incidents, together with a catalog of artificial intelligence flaws identified through reports, testing, evaluations, or investigations under this section. (B) Review and population of database Upon receipt of relevant information on an artificial intelligence security incident or artificial intelligence safety incident under paragraph (1), or on an artificial intelligence flaw under paragraph (1), the Director of the Institute shall review the information and determine whether the described incident or flaw constitutes an artificial intelligence security or artificial intelligence safety risk appropriate for inclusion in the database developed and established under subparagraph (A). (C) Identification of causal factors and artificial intelligence flaws When making a determination under subparagraph (B), the Director of the Institute shall identify causal factors for the artificial intelligence security incident or the artificial intelligence safety incident and determine whether the incident reveals, is associated with, or provides evidence of an artificial intelligence flaw. If the Director of the National Institute of Standards and Technology identifies such a flaw, the Director shall assign or associate the incident with a common flaw identifier and, to the extent practicable, identify other known incidents, near misses, models, or systems associated with the same flaw, including— (i) the artificial intelligence system; (ii) the deployment of the artificial intelligence systems; and (iii) practices related to the operation of the artificial intelligence system, including misuse of the artificial intelligence system. (3) Mandatory submissions (A) In general The following entities shall report any confirmed artificial intelligence safety incident or artificial intelligence security incident within 30 days of confirmation of such incident and within 72 hours if such incident poses an imminent threat to national security, critical infrastructure, or public safety: (i) Any developer or provider of a frontier artificial intelligence model. (ii) Any operator of critical infrastructure that utilizes an artificial intelligence model in the context of managing industrial control systems or other operational technologies. (B) Applicability For the purposes of subparagraph (A), the reporting requirement under this paragraph shall apply regardless of whether the frontier artificial intelligence model, or any model variant, checkpoint, or other version of such model involved in the incident, is or is intended to be made publicly available, and regardless of whether the incident occurs during development, training, testing, evaluation, red-teaming, deployment, or operation. (4) Priorities In evaluating information under paragraph (2) and determining under such subparagraph whether to include a report of an incident in the database required by paragraph (2)(A), the Director of the Institute shall prioritize inclusion in the database of cases in which a described incident— (A) describes an artificial intelligence system used in critical infrastructure or safety-critical systems; (B) would result in a high-severity or catastrophic impact to the people or economy of the United States; (C) includes an artificial intelligence system widely used in commercial or public sector contexts in the United States; or (D) constitutes a mandatory submission pursuant to subsection (a)(3). (5) Exemption from disclosure; reports and anonymity (A) Anonymity The Director of the Institute shall populate the database developed and established under paragraph (2)(A) with incidents and artificial intelligence flaws based on public reports and information shared using the mechanism established pursuant to paragraphs (1) and (3), ensuring that any incident description sufficiently anonymizes those affected, unless those who are affected have consented to their names being included in the database. (B) Exemption from disclosure Any information shared using the mechanism established pursuant to paragraphs (1) and (3)— (i) shall be exempt from disclosure and withheld, unless an affected party consents to the inclusion of their names in the database as provided for under subparagraph (A), from the public, pursuant to section 552(b)(3)(B) of title 5, United States Code, and any other provision of United States law or law of any State, political subdivision or agency thereof, or Tribe requiring disclosure of information or records; and (ii) shall not be deemed a waiver of any applicable privilege or protection, including trade secret protection. (C) Consultation required Before publishing information regarding an artificial intelligence safety incident or an artificial intelligence security incident, the Director of the Institute shall consult with the developer or provider of the artificial intelligence system involved in an incident. (b) Material risk guidance Not later than 180 days after the date of the enactment of this Act, the Director of the Institute shall, in coordination with the Director of the Cybersecurity and Infrastructure Security Agency, publish nonbinding guidance that provides illustrative criteria and examples for determining when an event materially increases a risk for purposes of an artificial intelligence safety incident or an artificial intelligence security incident.