HR-6461-119
Ordered to be Reported in the Nature of a Substitute (Amended) by the Yeas and Nays: 35 - 0.
Sponsored by Sarah McBride (D-DE)
What it does
This bill would direct the National Institute of Standards and Technology (NIST) to run a pilot program developing a standardized, modular template and technical guidelines for documenting artificial intelligence models. The template would capture basic information about AI models — such as the developer's identity, release date, training data cutoff, and supported languages — and would be available for voluntary use by both public and private sector entities. Within 12 months of launching the pilot, NIST would be required to report to Congress on the program's effectiveness and publish the template publicly on its website.
Who benefits
Businesses and government agencies that procure or deploy AI systems, who would gain a standardized way to evaluate what they are buying or using. Researchers and auditors who study AI systems, who would benefit from consistent, comparable documentation. Consumers and end users of AI-powered products, who may gain indirect transparency into the systems affecting them. Smaller AI developers who follow the voluntary standard, potentially gaining credibility and market access. International standards bodies and academic institutions involved in the collaborative development process.
Who is hurt
AI developers — particularly large companies with proprietary models — who may face competitive or reputational pressure to disclose information they consider trade secrets, even under a voluntary framework. Companies that have already invested in their own proprietary documentation formats may face costs to adapt to a new standard. NIST itself would bear administrative and staffing costs to run the pilot program, subject to appropriations. Foreign AI developers whose documentation practices differ from U.S. standards could face indirect market pressure if the template becomes a de facto industry requirement.
Supporters argue
Supporters argue that the rapid proliferation of AI systems in high-stakes domains — hiring, healthcare, criminal justice, and financial services — has outpaced the public's ability to understand what these systems are and how they work. They contend that a voluntary, NIST-developed documentation standard, modeled on NIST's successful cybersecurity framework, would give buyers, regulators, and the public a consistent baseline of information without imposing mandates on innovators. They further argue that the bill's collaborative, stakeholder-driven process — including a 60-day public comment period and input from industry, academia, and international standards bodies — ensures the resulting template will be practical and broadly accepted.
Opponents argue
Opponents argue that a voluntary documentation standard without enforcement mechanisms is unlikely to produce meaningful transparency, since developers with the most to hide face no obligation to participate. They contend that NIST's existing AI Risk Management Framework already addresses documentation practices, making this pilot program potentially duplicative and an inefficient use of limited agency resources. Critics also argue that a government-designed template may quickly become outdated given the pace of AI development, and that the bill's open-ended scope — covering all AI models in both public and private sectors — could create confusion about what disclosures are expected versus required.