Stop Rogue AI Act
- Introduced
- Passed House
- Passed Senate
- Became Law
Referred to committee
Would set federal standards that give AI agents a verified identity, constant monitoring, permanent logs and an off switch, and write them into federal contracts. Agents that people build on open models and run on their own machines could not meet them.
Latest update
Introduced
Take action
This bill is still moving, so there's still time to push back.
Contact CongressTake action on this issueThe bill
Read the full billOfficial bill recordKey facts
- Where
- United States
- Bill number
- H.R. 10362 (119th)
- Status
- In progress
- Introduced
- September 14, 2026
- Threat
- Medium
- Last update
- September 14, 2026
Topics
Share this bill
What it does
H.R. 10362 directs NIST, working with the NTIA, to publish standards for AI agents within a year and update them annually. Under those standards, organizations using AI agents would keep a live, machine-readable inventory of all of them; verify an agent's identity with cryptography at the network and application level, without relying on the agent's own claim; monitor agents in real time and intercept their actions where needed; trace an agent to the person or company that built or runs it; and keep tamper-evident logs of what agents do. NIST must promote DNS-based identity and agent registries, and the standards must be open and not tied to any one company's registry. Within 18 months of the standards, the Federal Acquisition Regulatory Council must propose rules requiring federal agencies and contractors to follow them, including the power to allow, deny or limit what agents do. Contracts signed before the law takes effect are exempt. OMB and CISA would then issue guidance to agencies. The bill sets no direct rules for private companies or individuals.
What’s at stake
An agent built on an open-weight model and run on someone's own computer has no company behind it to vouch for its identity, and the bill says an agent's own claim is not enough. Under these standards it fails by default. The binding rules cover only federal agencies and contractors, but NIST standards tend to become the industry default: large companies adopt them, then demand them from partners and users. Services that follow them would admit registered, traceable agents and refuse the rest. Open models and privately run agents would still exist, shut out of the services people use most. The push for agent registries also creates a gatekeeper that decides which agents count as legitimate.
Our take
We oppose H.R. 10362. The published text confirms what the sponsors announced: a federal standard for AI agent identity, logging and shutdown. Organizations can already inventory software and block what they don't trust on their own networks. A national standard turns traceable identity into the price of access and pushes open-weight and locally run AI to the margins. The bill's demand for open standards with no single owner lowers the risk of one company acting as gatekeeper. It does nothing for a person who wants to run an AI agent privately, with no registry and no logs.
Timeline
IntroducedLatest
Referred to committee
Gottheimer and Lawler announced the bill, five days before it was formally introduced
The record
Sponsor
Josh Gottheimer (D-NJ)
Cosponsors (2)
Michael Lawler (R-NY), Jared Huffman (D-CA)
Committee
Oversight and Government Reform Committee, Science, Space, and Technology Committee
Suggest a bill or correction
Know a bill we should track, or spot something out of date? Email the bill name, the jurisdiction and a link to an official source.