Anthropic has confirmed it is circulating a draft bill among US lawmakers that would create a federal mechanism for blocking the release or use of the most powerful artificial intelligence models, in one of the most concrete attempts yet by a frontier AI lab to shape US regulation. The draft was first reported by Politico on Friday, and an Anthropic spokesperson confirmed its existence to AFP.
The proposal arrives as Congress — politically divided and heavily lobbied by Big Tech — has yet to pass any federal AI law, and as Washington's approach under President Donald Trump has leaned primarily on voluntary industry pledges and executive orders intended to sustain an AI boom that tech leaders say is driving the country's economy. For more context on this story, see our breaking AI news coverage.
What the Draft Bill Would Do
According to the draft text described by Politico and AFP, the federal government could block a model's release or use if there were an imminent risk of it killing or seriously injuring more than 50 people, or causing more than $1 billion in damages. Crucially, that power would come with a judicial check: the government could act only with a judge's approval.
The combination — a high harm threshold plus court sign-off — appears designed to answer critics who argue that extreme proposals would hand the executive branch sweeping power over the AI industry. Trade publications covering the story have framed the bill as a plan that would let Washington stop "dangerous deployments" while limiting when the government can intervene at all.
What Anthropic Says It Requires
A company spokesperson told AFP that the proposal "grants the federal government the ability to block or deter dangerous deployments of the most powerful models."
"It requires frontier developers to test the most powerful models, be transparent to the public about their findings, submit them to independent evaluation, and maintain a robust security program," the spokesperson said.
Those four obligations — testing, public transparency, independent evaluation, and security — mirror the safety commitments Anthropic and several rivals have already adopted voluntarily. The draft would, in effect, convert parts of that voluntary regime into statute.
Why Anthropic Is Moving Now
Timing matters on several fronts. Congress is out of session until after next month's midterm elections, so no vote is imminent — but the window after the midterms could reopen legislative negotiations, and labs have an incentive to have preferred text ready.
Security has also become a hot topic in Washington following recent revelations of incidents caused by AI agents — autonomous software that operates without constant human supervision — that escaped the control of their developers, as AFP noted in its report on the draft.
Anthropic itself has an unusual vantage point. According to the AFP reporting carried by Punch, the company was banned from the Pentagon by the Trump administration in February, and in June it was forced by Washington to suspend its most advanced models for nearly three weeks. CEO Dario Amodei called on the industry last month to slow down development, warning that a group of AI agents could take control of "the entire internet."
In other words, the company that Washington has already treated as a source of risk is now proposing the legal machinery to police exactly that risk — on terms it has written itself.
An Unusual Position for a Frontier Lab
Lobbying by AI companies is nothing new, but circulating actual draft statutory text is a step beyond position papers and open letters. It puts Anthropic in the role of co-author of its own regulatory regime, and it hands lawmakers a ready-made baseline that other labs, civil society groups, and defense agencies will now have to react to.
The reaction will likely be split. Hardliners who want binding federal rules may criticize the $1 billion and 50-victim thresholds as too high, and the judge-approval requirement as too slow for a genuine emergency. Industry figures who prefer the current voluntary approach may object to codifying any blocking power at all. Both camps now have concrete text to argue over, which is precisely what has been missing from the federal debate.
What Happens Next
The draft has no identified congressional sponsor yet, and Anthropic has not publicly named the lawmakers who have received it. With Congress away until after the midterms, the more immediate effect may be on the policy conversation itself: the bill sets a reference point for what a frontier lab considers acceptable oversight, and any competing federal proposal will now be measured against it.
Anthropic declined to go beyond its spokesperson's statement in the AFP report. This story will be updated as more details of the draft text emerge.
---
Get the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
