Daily Agentic, the front page of the machine age
Today
Policy & Rules / Anthropicthe 23rd of October 2026, 15:30 UTC

Anthropic bill lets judges halt AI: 50 lives

Anthropic's draft US law would let a judge stop an AI model that puts more than 50 lives or 1 billion dollars at risk, TechCrunch reported on the 20th of October 2026. Critics call the bar too high, and Dario Amodei answered them on X.

Anthropic's draft AI law: judge decides

Anthropic, maker of the Claude chatbot, wants a US law to stop dangerous AI. A judge could halt an AI that risks over 50 lives or 1 billion dollars, TechCrunch reported on the 20th of October 2026.

AI safety campaigners on X call that bar too high. Anthropic's boss, Dario Amodei, answered a critic on X, once called Twitter, on the 21st of October.

Below we break down who could stop an AI, and when

It is a draft bill, a text Congress can make law. Politico broke the news on the 16th of October, and Anthropic said it was real, TechCrunch wrote.

Outside testers would check the strongest AI. They would sign up with the Commerce Department, the US office for business.

If they find a near, grave danger, they must warn Commerce in 72 hours, TechCrunch wrote. Commerce then asks a judge in Washington to stop the AI.

The draft rule in five lines: a judge can stop an AI at a risk of more than 50 deaths or serious injuries, or more than 1 billion dollars in damage; outside testers listed with the Commerce Department check it; they have 72 hours to warn; a judge in Washington gives the stop orderThe draft rule in five lines: a judge can stop an AI at a risk of more than 50 deaths or serious injuries, or more than 1 billion dollars in damage; outside testers listed with the Commerce Department check it; they have 72 hours to warn; a judge in Washington gives the stop order

The danger must come from a germ weapon, a computer attack, or an AI that escapes its maker, TechCrunch wrote.

Anthropic: "We want a court to act quickly"

"We want a court to act quickly in an emergency," Anthony Cimino wrote on X on the 18th of October. He leads Anthropic's work with Congress.

The 50 lives rule comes from a California law, he wrote. He wrote that Anthropic plans to change the part that overrules state laws.

A critic, and Amodei's four months on Mythos

Nathan Calvin works at Encode AI, a group that wants safety rules for AI. On the 17th he wrote on X that the bill is "not remotely close to meeting the moment."

Amodei's reply named nine safety steps that cost his firm money. One was holding back Mythos, a family of AI models.

Anthropic had Mythos ready in February, he wrote. Until June, only a few firms that guard computers could use it.

"We likely sacrificed a significant chunk of a several-month model lead," he wrote.

What is confirmed and what is not

The bill text is still private. Anthropic's news page has no post on it.

Some news sites date the post the 20th, its US date.

Key facts

Politico broke the news on the 16th of October 2026.

Only a judge could stop an AI that risks over 50 deaths.

Mythos was held back four months, Amodei wrote.

Source: TechCrunch, reporting Politico; Anthropic's Anthony Cimino and Dario Amodei on X