Secure A.I. Development Act of 2026

This bill would create an Artificial Intelligence Risk Board within the National Institute of Standards and Technology (NIST) to develop tests and best practices for advanced, or "frontier," AI models. Companies would have to give the National Security Agency access to a frontier AI model at least 21 days before releasing it and register the model in a federal registry, with fines of at least $100,000 per day for violations of the NSA access rule. NIST would set up a voluntary system for reporting AI security and safety incidents and a public database of those incidents. Federal agencies would update how they track and handle computer security weaknesses related to AI, and the NSA would run a pilot program of at least 3 years to share threat information with AI companies that serve the federal government.

Full text and official record on Congress.gov

Sponsors

• Mark R. Warner Democrat sponsor
• 0 cosponsors