# Anthropic CEO Takes Responsibility as Draft AI Legislation Draws Criticism

By Riza Dagoc

Canonical URL: https://www.tokenpost.com/news/regulation/40819
Published: 2026-10-26T06:00:16.000Z
Updated: 2026-10-26T06:00:16.000Z
Section: Regulation

> Amodei said some criticized actions could have been handled better. Anthropic plans to revise its state-preemption language but has not set a date.

Anthropic CEO Dario Amodei accepted responsibility for the company’s proposed federal AI safety legislation after criticism of its model-pause and state-preemption provisions.

“I am of course the CEO, and I take responsibility,” Amodei wrote Oct. 20 in a reply to critic Nathan Calvin. Amodei said some actions Calvin criticized, including the draft legislation, could have been handled better.

Anthropic federal affairs chief Anthony Cimino said Oct. 18 that the company had turned its June AI safety framework into draft legislation and shared it with a small number of congressional offices. Cimino said Anthropic planned to revise the language on preemption of state laws but gave no deadline.

Anthropic has not published the draft text. Under the proposal, only the Department of Commerce could ask a federal court in Washington, D.C., to pause a model. Independent evaluators would have 72 hours after completing an assessment to report an imminent covered risk to the department. The reported trigger covers a foreseeable, imminent risk involving a biological weapon, a model evading its developer’s controls or a cyberattack, with potential for more than $1 billion in damages or death or serious injury to more than 50 people.

AI safety writer Zvi Mowshowitz argued that the proposal combines stronger preemption of state laws with weaker enforcement than the FRONTIER Act. California state Sen. Scott Wiener said it “falls far short of protecting public health and safety.” Anthropic’s June framework calls for government authority to block dangerous deployments and says federal law should not preempt state law unless it is at least as strong as the framework. The framework lists four risk areas, including automated AI research and development.

The same numerical cutoffs appear in the FRONTIER Act, H.R. 9925, introduced July 23. It defines the threshold as a risk of death or serious injury to more than 50 people, or more than $1 billion in damage to or loss of property. The act would let the secretary of Commerce issue a provisional emergency order after a preliminary risk finding; the order would lapse 45 days after service, or earlier in specified circumstances.