The FRONTIER Act would put federal oversight on frontier AI — defining catastrophic risk numerically at more than 50 casualties or a billion dollars
H.R.9925, the FRONTIER Act, provides for federal oversight of the development and deployment of frontier artificial intelligence in interstate and foreign commerce. What distinguishes it is a numerical definition of catastrophic risk: a foreseeable and material risk of contributing to the death of or serious injury to more than 50 people, or more than a billion dollars in property damage or loss, from a single incident.
Bill overview (primary data)
- Bill numberH.R. 9925
- TypeHouse Bill
- Congress119th Congress
- Latest actionReferred to the Committee on Energy and Commerce, and in addition to the Committee on Science, Space, and Technology, for a period to be subsequently determined by the Speaker, in each case for consideration of such provisions as fall within the jurisdiction of the committee concerned.（2026-07-23）
Key points
- H.R.9925, the FRONTIER Act, provides for federal oversight of frontier AI development and deployment in interstate and foreign commerce.
- Catastrophic risk is defined numerically: more than 50 deaths or serious injuries, or more than $1 billion in property damage or loss, from a single incident.
- Contemplated incidents involve weapons assistance not publicly available, and cyberattack or criminal-equivalent conduct with no meaningful human oversight.
- Acceptable mitigation means enough to ensure anticipated benefits outweigh the level of catastrophic risk — a comparative judgment, not zero risk.
- Licensed independent verification organizations assess very large frontier developers frameworks and practices.
1Defining catastrophic with a number
The hardest part of writing AI regulation is deciding where the regulated territory begins. This bill gives the phrase catastrophic risk a clear numerical threshold: death or serious injury to more than 50 people, or more than a billion dollars in property damage or loss, arising from a single incident. Not a vague adjective but a line drawn before the argument starts.
2The pathways it contemplates
- 1Weapons assistanceProviding assistance not publicly available in the development or release of a chemical, biological, radiological, nuclear or cyber weapon
- 2Autonomous cyberattackEngaging in a cyberattack with no meaningful human oversight or intervention
- 3Autonomous criminal conductEngaging, with no meaningful human oversight or intervention, in conduct that would constitute murder, assault, extortion or theft if committed by an individual
All three carry the condition of no meaningful human oversight or intervention. What is at issue is behavior arising from the model itself rather than a result a person directed. On weapons, the qualifier assistance not publicly available draws a line so that summarizing openly available information does not qualify.
3A standard set against benefits
The definition of acceptable levels of catastrophic risk mitigation deserves attention too. It does not demand that risk be reduced to zero; it means mitigation adequate to ensure the anticipated benefits of a frontier model outweigh its level of catastrophic risk, taking probability and magnitude into account.
The design calls for a comparative judgment rather than absolute safety, and the stance of building regulation while acknowledging the technology benefits shows up already in the definitions.
4Independent evaluation as a mechanism
The bill defines an assessment as a review conducted by an independent verification organization licensed under the act, examining whether a very large frontier developer frontier AI framework, governance policies and practices, risk-monitoring and mitigation of detected risks are adequate to achieve acceptable levels of catastrophic risk mitigation.
Rather than the executive reviewing directly, licensed third parties evaluate. Of the 120 bills this site holds as of 2026-09-02, 94 (78 percent) remain referred to committee, and this bill is at that stage.
Why it matters
For frontier model developers, whether the regulatory threshold is expressed numerically changes predictability substantially. Assessment by licensed independent organizations is also worth noting as a form of oversight distinct from direct review by the executive.
FAQ
Why define it numerically?
Would summarizing public information qualify?
Sources (primary)
Source: Congress.gov (Library of Congress; U.S. legislative materials, public domain). Links go to the official site.
- Congress.gov (bill page, original)
- H.R. 9925（119th Congress）