VET Artificial Intelligence Act
- Sponsor
- Sen. Hickenlooper, John W. [D-CO]
- Committees
- Commerce, Science, and Transportation Committee (primary)
- Last action
- Jul 31, 2025
Bottom line
This bill aims to foster trust and accountability in AI systems by standardizing voluntary assurance practices and evaluating the ecosystem of AI assurance providers.
What it actually does
This bill mandates the National Institute of Standards and Technology (NIST) to develop and regularly update voluntary technical guidelines and specifications for both internal and external assurance of artificial intelligence (AI) systems. It also establishes an advisory committee to recommend qualifications for AI assurance professionals and requires a study on the capabilities of the AI assurance sector.
Proponents argue
Supporters argue that establishing clear, voluntary guidelines for AI assurance will enhance trust, promote responsible development and deployment of AI, and provide a framework for accountability, thereby accelerating AI adoption and mitigating potential harms. They believe it aligns with existing NIST efforts like the AI Risk Management Framework.
Opponents contend
Critics might contend that while well-intentioned, voluntary guidelines may not be sufficient to address the rapid evolution and potential risks of AI, potentially leading to a patchwork of compliance or insufficient oversight. Some may also argue that mandating such guidelines could stifle innovation or impose undue burdens on developers and deployers, particularly smaller entities, if not carefully implemented.
The bill is relatively concise and its structure allows for a clear understanding of its main objectives within a reasonable reading time.
Section 4
NIST Guidelines for AI Assurance
This section requires the Director of the National Institute of Standards and Technology (NIST) to develop and regularly update a set of voluntary technical guidelines and specifications for internal and external artificial intelligence (AI) assurance. These guidelines will cover safeguards for consumer privacy, methods to assess and mitigate harms, dataset quality, documentation, disclosure, and governance, aiming to standardize testing, evaluation, verification, and validation methods.
Supporters argue
Supporters argue that these guidelines are crucial for establishing a common understanding of what constitutes trustworthy AI, fostering responsible innovation, and providing a mechanism for organizations to demonstrate their commitment to ethical and safe AI practices. They emphasize the voluntary nature allows flexibility while still promoting best practices.
Critics contend
Critics might express concern that voluntary guidelines may not be sufficient to ensure compliance, especially for high-risk AI applications, and could lead to a 'check-the-box' mentality without genuine accountability. Some may also argue that the rapid pace of AI development means any fixed guidelines could quickly become outdated.
Tradeoffs
The tension lies between the desire for flexible, voluntary adoption to encourage innovation and the need for robust, enforceable standards to ensure safety and ethical use of AI. The bill attempts to balance this by making the guidelines voluntary but comprehensive.
Section 5
Establishment of AI Assurance Qualifications Advisory Committee
This section directs the Secretary of Commerce to establish an advisory committee, comprising up to 20 experts from various sectors including academia, industry, consumer groups, and labor organizations. The committee's role is to review case studies and make recommendations on the qualifications, expertise, licensing, independence, and accountability required for parties conducting AI system assurances, considering both internal and external processes.
Section 6
Assessment of the AI Assurance Sector
This section requires the Secretary of Commerce to conduct a study evaluating the capabilities of entities that perform internal and external AI assurances. The study will assess their personnel, technical tools, evaluation methods, computing and physical infrastructure, market demand, and the feasibility of leveraging existing accreditation programs like NIST's National Voluntary Laboratory Accreditation Program (NVLAP) for external AI assurances. A report with findings and recommendations will be submitted to Congress.
Defines 'nonaffiliated third party' for external AI assurance, requiring no common ownership or corporate control, financial independence, no shared employees with the developer/deployer, and demonstrated expertise in relevant technical domains (data privacy, security, risk management) and familiarity with the AI system type.
Section 3(7)
Why it matters:This is standard legislative practice to define terms precisely at the beginning of an act to ensure consistent interpretation throughout the bill. It's not evasive.
Case for: Supporters argue that a strict definition of 'nonaffiliated third party' is essential to ensure the credibility, impartiality, and trustworthiness of external AI assurance, preventing conflicts of interest and guaranteeing expert evaluation.
Case against: Some might argue that the criteria, particularly regarding financial independence and no shared employees, could be overly restrictive, potentially limiting the pool of qualified third-party evaluators, especially for specialized AI systems or smaller developers.
Estimated impact: This definition will shape the market for external AI assurance services, influencing which entities can provide these services and the level of independence required.