S. 2615 — VET Artificial Intelligence Act

This section states the purposes of the Act: (1) developing consensus-driven, evidence-based voluntary technical guidelines and specifications for internal and external assurances of artificial intelligence systems through testing, evaluation, validation, and verification, as appropriate based on the system's intended application, use-case, and risk profile; (2) using such assurances to supplement methodologies for building trust in, increasing adoption of, and providing accountability and governance for artificial intelligence systems; and (3) furthering the goals of the Artificial Intelligence Risk Management Framework, including any successor, published by the National Institute of Standards and Technology and the Artificial Intelligence Safety Institute pursuant to section 22A(c) of the National Institute of Standards and Technology Act (15 U.S.C. 278h–1(c)).

This section establishes definitions for terms used in the Act, including (1) "artificial intelligence system," meaning a machine-based system that, for explicit or implicit objectives, infers from inputs how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments; (2) "deployer," meaning an entity that operates an artificial intelligence system for internal use or for use by a third party; (3) "developer," meaning an entity that builds, designs, codes, produces, trains, or owns an artificial intelligence system for internal use or for use by a third party (excluding an entity that is solely a deployer); (4) "Director," meaning the Director of the National Institute of Standards and Technology; (5) "external artificial intelligence assurance," meaning an independent evaluation of an artificial intelligence system by a nonaffiliated third party using specified guidelines or standards to verify functionality, fitness for purpose, or identify errors, vulnerabilities, or negative societal impacts; (6) "internal artificial intelligence assurance," meaning an independent evaluation of an artificial intelligence system by the evaluated party using an impartial internal structure to verify functionality, fitness for purpose, or identify errors or vulnerabilities; (7) "nonaffiliated third party," meaning a financially independent, qualified evaluator of artificial intelligence systems lacking common ownership, control, or employment ties to the developer or deployer and possessing relevant expertise in data privacy, security, and risk management; and (8) "Secretary," meaning the Secretary of Commerce.

This section directs the Director of the National Institute of Standards and Technology (NIST), in collaboration with public and private sector organizations including the National Science Foundation and Department of Energy, to develop voluntary technical guidelines and specifications for internal artificial intelligence (AI) assurance (i.e., developer-led) and external AI assurance (i.e., third-party) within one year of enactment, with review and updates at least every two years. The guidelines must (1) identify consensus standards addressing consumer privacy safeguards, harm mitigation methods, dataset quality, documentation and disclosure, and governance controls; (2) provide best practices and common definitions for testing, evaluation, verification, and validation; (3) recommend criteria for assurance frequency, scope, disclosure of results, and corrective actions based on AI system risk and use case; (4) align with voluntary consensus and international standards and industry best practices; and (5) protect confidentiality of sensitive information without requiring specific solutions or technologies. It further requires NIST to solicit at least 30 days of public comment on drafts, make drafts publicly available, convene stakeholder forums, and publish the final guidelines on its website.

This section directs the Secretary to establish, not later than 90 days after publication of voluntary AI technical guidelines under section 4(a), the Artificial Intelligence Assurance Qualifications Advisory Committee with not more than 20 members possessing AI expertise, including at least one representative from each of 12 categories (e.g., institutions of higher education, AI developers and deployers, civil rights organizations, labor organizations). The committee must review case studies on licensure, certification, and accreditation for compliance verification; assess their applicability to AI systems; and, within one year of establishment, submit to the Secretary and Congress (and publicly release) recommendations on qualifications, licensing, independence, and accountability for AI assurance providers—considering AI type and internal/external processes—and on using existing programs for such accreditation. The committee terminates one year after submitting its recommendations.

This section directs the Secretary to commence, not later than 90 days after publication of voluntary technical guidelines under section 4(a), a study evaluating capabilities of entities conducting internal artificial intelligence (AI) assurances (i.e., by developers or deployers) and external AI assurances (i.e., by third parties). The study must (1) assess adequacy of such entities' personnel, technical tools, evaluation methods, computing infrastructure, and physical infrastructure for assurances compliant with the guidelines; (2) review features, best practices, and safeguards for maintaining integrity of confidential or proprietary information; (3) evaluate market demand for and availability of such entities; and (4) assess feasibility of leveraging facilities accredited under the National Voluntary Laboratory Accreditation Program for external AI assurances. Not later than one year after commencing the study, the Secretary must submit a report with study results—including recommendations to improve entity capabilities and availability, descriptions of assessed features and safeguards, and conclusions on accredited facilities—to the Senate Committee on Commerce, Science, and Transportation; the House Committee on Science, Space, and Technology; and relevant federal agency heads.