Congress is weighing a growing stack of proposals to oversee the most powerful artificial intelligence systems. Lawmakers, the White House, state governments and the industry itself still disagree on how much control government should have over so-called frontier AI models, and who should wield it. The reporting reviewed for this article describes the bills as proposed or under negotiation, not enacted.
Key Highlights
- Several Senate and House bills would set up testing, reporting or review requirements for the most capable AI models.
- A confidential industry draft reported by Politico would let the government block a model only in narrow, catastrophic cases.
- The Trump administration opposes mandatory AI regulation and favors voluntary, industry-developed standards, according to PYMNTS.
- Whether federal rules should override state AI laws remains a central dispute.
What Lawmakers Have Proposed
Sens. Mark Warner (D-Va.), Brian Schatz (D-Hawaii) and Andy Kim (D-N.J.) introduced the Artificial Intelligence Risk Management and Security Act of 2026, PYMNTS reported on October 12. The bill would create an AI Safety Board inside the Commerce Department to evaluate models that could threaten national and economic security. The text is available on govinfo.gov.
In a press release, Warner said that if a model can find and exploit weaknesses in banks, water systems or the power grid, the public should know before it is released, not after something goes wrong.
Other proposals are circulating, according to PYMNTS:
- Sen. Maria Cantwell (D-Wash.) offered a framework directing the National Institute of Standards and Technology to develop standards against catastrophic harms such as AI-enabled cyberattacks and biological threats.
- Sen. Josh Hawley (R-Mo.) proposed the AI Risk Evaluation Act, which would have the Department of Energy stress-test top models before release and penalize companies that skip review.
In the House, the FRONTIER Act, led by Reps. Lori Trahan (D-Mass.) and Jay Obernolte (R-Calif.), would give the government power to stop deployment of risky models and override some state laws, according to Politico’s reporting. In the Senate, Amy Klobuchar (D-Minn.) and Ted Cruz (R-Texas) are negotiating a bill aimed at catastrophic risks.
The Reported Anthropic Draft
Politico reported on October 16, citing three people with knowledge who spoke anonymously, that AI company Anthropic has quietly pitched lawmakers a draft frontier AI safety bill. Anthropic has not introduced the bill, and its contents rest on that reporting. An Anthropic spokesperson confirmed the draft to Politico, according to TechCrunch.
According to Benzinga’s summary of the Politico report, the draft would allow a model to be blocked only if it posed an imminent risk of more than $1 billion in damage or more than 50 deaths or serious injuries. The risk would have to come from a biological weapon, a model escaping its developer’s controls or a cyberattack. Only a federal judge in Washington, acting on a Commerce Department request, could suspend a model.
The draft would also use independent evaluators, who would report imminent risks to Commerce within 72 hours. It would preempt some state laws, and Politico noted Democrats have resisted such proposals. An Anthropic spokesperson told Politico, “These are sensible policies we believe Congress should act on.”
Can Governments Control the Risks?
Supporters of federal rules argue that testing before release is the only way to catch dangerous capabilities early. Critics of broad powers worry about regulators’ discretion. Coverage by Interesting Engineering describes the debate as turning on whether one national framework can protect the public without giving regulators too much latitude.
States are already acting. The Future of Privacy Forum noted in an analysis of federal and state proposals that Illinois passed a frontier model safety bill and that the White House issued an executive order on AI security. It also cautioned that federal AI legislation has faced a difficult path in Congress.
The White House is a major obstacle. PYMNTS reports that the Trump administration remains adamantly opposed to mandatory AI regulation, though it said opposition may not be enough to stop efforts backed by the public, industry and many statehouses.
What Remains Uncertain
It is not yet clear which bill, if any, will move forward. Control of Congress after the November 3 midterms could shape the outcome: PYMNTS noted Warner could gain a committee gavel if Democrats take control. Details of the Anthropic draft could also change, since it has not been formally released.
What to Watch Next
- Whether any bill is formally introduced or scheduled for a vote.
- How negotiators handle the preemption of state laws.
- Any formal White House response to the proposals.
Sources: PYMNTS (Oct. 12), Politico as summarized by Benzinga (Oct. 16), TechCrunch, Interesting Engineering, Future of Privacy Forum, Senate bill text on govinfo.gov. This article will be updated as new information is confirmed.
Disclosure: This article was prepared with the assistance of Claude, an AI assistant developed by Anthropic, the company discussed in this report. Our review process is described in our Fact-Checking Policy.