Stop Rogue AI Act
United States- Introduced
- Passed House
- Passed Senate
- Became Law
Would turn verified identity, continuous inventories, real-time monitoring, persistent logs, and revocable access into federal standards for AI agents, creating a control system that could shut open-weight and privately run agents out of trusted networks.
Latest update
Gottheimer and Lawler announced the bill as introduced
This bill is still moving
This bill is still moving, so there's still time to push back.
Contact LawmakersWhat it does
According to official sponsor materials, the Stop Rogue AI Act directs NIST to develop standards for discovering and controlling AI agents. The described standards include a continuous machine-readable inventory, cryptographically verifiable identity and provenance at the network and application layers, real-time monitoring, tamper-evident logs, and controls to allow, deny, constrain, or revoke agent activity. Federal agencies and contractors procuring or deploying agents would be required to follow the resulting standards. Both sponsors say the measure was introduced on September 9, 2026, but no bill number or official legislative text was publicly available when this entry was checked.
What’s at stake
Open-weight agents can be copied, modified, and run locally without a central vendor continuously verifying who built or operates them. Making recognized identity, provenance, monitoring, and portable logs the price of being trusted would place anonymous, pseudonymous, community-built, and privately run agents at a structural disadvantage. The immediate mandate is described as applying to federal agencies and contractors, but federal procurement standards often shape the wider market. The result could be an AI ecosystem in which access depends on traceable identity and continuous surveillance, while software that preserves user privacy is blocked as unverifiable.
Our take
We oppose the bill as the sponsors currently describe it. Organizations already have authority to inventory software and block unauthorized activity on their own networks. A federal identity and traceability standard expands that local security decision into system-wide control. Open weights let users run software beyond a central provider’s monitoring, logging, and revocation systems; that is a privacy strength. Security standards should enforce network boundaries while preserving anonymous, private, and locally controlled AI. This assessment must be revisited when the actual legislative text becomes public.
Timeline
Gottheimer and Lawler announced the bill as introduced
LatestSponsors
Josh Gottheimer (D-NJ), Mike Lawler (R-NY)
Suggest a bill or correction
Is there a bill we should include, an update we’ve missed, or a detail that needs correcting? Email us with the bill name, jurisdiction, and a link to an official source if you have one.