FRONTIER Act

This bill creates federal oversight of advanced AI systems called frontier models, trained with very large amounts of computing power, through a new Under Secretary of Commerce for AI Security. Larger developers would have to publish a safety framework for managing catastrophic risks, get an annual third-party compliance audit, register with the Under Secretary, and pay fees. All frontier developers would have to publish transparency reports when releasing models and report critical safety incidents within set deadlines. The largest developers would have to hire a federally licensed independent verification organization to assess their safety practices at least every 6 months, and those organizations would be largely shielded from lawsuits. The Secretary of Commerce could issue emergency orders to suspend or restrict a model that poses an imminent catastrophic risk, and violations could bring civil penalties enforced by federal and participating state attorneys general.

Full text and official record on Congress.gov

Sponsors

• Jay Obernolte Republican sponsor
• 11 cosponsors