Anthropic representatives have been quietly pitching Capitol Hill lawmakers on a draft bill to regulate the safety and security of frontier AI models, according to a Friday report — one that would, in narrow extreme cases, let the government stop a model from being released or used. Why it matters: an AI lab is asking Congress to write its own industry into law, with specific thresholds and outside evaluation attached.
October 16, 2026
What happened
Politico reported on Friday that Anthropic has shared draft legislation with lawmakers, citing three people familiar with the effort. An Anthropic spokesperson later told Reuters the draft reflects proposals the company set out earlier this year in its Advanced AI Framework, and described its purpose as giving the government a way to block or deter dangerous deployments of the most powerful models. The Department of Commerce did not immediately comment. This is a pitch, not a filed bill: no sponsor has been named and no vote is scheduled. What is new is the specificity — actual draft language, with numbers in it.
How the draft would work
Under the reported draft, the government could move to block a model’s release or use only when it poses a foreseeable and imminent risk above a high bar: more than $1 billion in damages, or more than 50 people killed or seriously injured. The risk would also have to fall into defined categories, such as helping create a biological weapon, a model escaping its developer’s controls, or a cyberattack.
Independent evaluators would do much of the checking. They would register with the Department of Commerce and periodically assess a lab and its frontier models. Labs could pay for evaluations, as long as payment was not tied to the outcome. If an evaluator found an imminent risk of the kind listed in the bill, it would have to report that finding to Commerce within 72 hours of finishing its assessment.
If Commerce concluded a model met the risk test, it could ask a judge to halt it, with a federal judge in the District of Columbia able to order a suspension. The draft would bar any other federal department, agency or official from suspending a model’s development, deployment or internal use on their own. The spokesperson’s description added broader duties for developers: testing frontier models, being transparent about findings, and submitting to independent evaluation.
Why it matters
For people who use AI tools every day, the practical point is the shape of the compromise. The draft sets a deliberately high threshold for intervention — imminent, large-scale harm — while making testing and outside evaluation routine rather than voluntary, and concentrates blocking power in one department and one court. It lands in a busy week for AI oversight, after regulatory demands sent to major labs and overseas hearings where Anthropic backed mandatory rules. Whether Congress takes up this draft, a competing one, or nothing at all is now the open question.
Frequently asked questions
Has the bill been introduced in Congress?
Not according to the reporting. Anthropic has pitched draft language to lawmakers; no sponsor or formal introduction was reported.
When could the government block a model under it?
Only for a foreseeable, imminent risk above $1 billion in damages or more than 50 people killed or seriously injured, in listed categories such as bioweapon creation, loss of developer control, or a cyberattack — and a federal judge would have to order the suspension.
Who would check frontier models?
Independent evaluators registered with the Department of Commerce, paid by the labs but not based on outcomes, with a duty to report an imminent listed risk to Commerce within 72 hours.
Sources: Reuters, Politico