FRONTIER Act

The FRONTIER Act establishes federal oversight for the development and deployment of frontier artificial intelligence models in interstate and foreign commerce. It requires large frontier developers to create and publish risk management frameworks, mandates independent verification for very large developers, and gives the Secretary authority to issue emergency orders for imminent catastrophic risks, while preempting certain state laws on AI risk transparency, auditing, and incident reporting.

Sec. 2, Definitions: Defines key terms including 'catastrophic risk' as a material risk of death or serious injury to over 50 people or over $1 billion in property damage from a single incident involving a frontier model, such as assisting in weapon development, conducting cyberattacks without human oversight, or evading controls. Also defines 'critical safety incident' as unauthorized access to model weights, harm from catastrophic risk, loss of control, or deceptive subversion by the model.

Sec. 3, Rulemaking: Directs the Under Secretary to issue regulations within 180 days establishing minimum requirements for frontier AI frameworks, which must be reviewed annually. Also requires regulations for licensing and oversight of independent verification organizations (IVOs), including conflict-of-interest rules, application procedures, and grounds for license revocation.

Sec. 4, Transparency and Reporting: Requires large frontier developers to write, implement, and publish a frontier AI framework within one year or 90 days after qualifying. The framework must cover risk thresholds, assessments, cybersecurity for model weights, critical safety incident response, and internal governance. Developers must update the framework annually and publish justifications for material modifications, and retain a third party for annual compliance audits.

Sec. 5, Independent Verification: Requires very large frontier developers to retain a licensed IVO for ongoing assessment of their framework and risk mitigation. IVOs must be licensed by the Under Secretary, may subcontract portions of assessments but remain fully responsible, and must be granted timely access to materials. Assessment reports must be submitted with a frequency based on risk severity.

Sec. 6, Cumulative Obligations: Large frontier developers must comply with all applicable requirements, and very large frontier developers must comply with requirements for both frontier and large frontier developers.

Sec. 7, GAO Report on IVO Market: Requires the Comptroller General to submit an annual report to Congress on the IVO market, including the number of licensed IVOs, their capacity, barriers to entry, independence from the AI industry, and recommendations for policy changes.

Sec. 8, Emergency Orders: Authorizes the Secretary to issue emergency orders suspending or restricting development, deployment, or internal use of a frontier model if it presents an imminent catastrophic risk. Orders must include factual basis, technical assessment, and criteria for rescission. The Secretary has final authority after consulting the Under Secretary.

Sec. 9, Relationship to State Laws: Preempts states from adopting laws that impose new substantive obligations on AI developers regarding frontier AI risk transparency, third-party auditing, and incident reporting. Does not affect generally applicable laws, state regulation of AI deployers or users, or other specified state authorities.