Anthropic Circulates Draft AI Safety Bill Among US Lawmakers
publishers 128articles 219first reported 24 Sep, 16:43 UTCdeveloping since 24 Sep · 34 editionstop 25 #19
Anthropic is circulating draft legislation among US lawmakers that would establish federal safety and security rules for the most powerful artificial intelligence models, the company confirmed Friday.
Under the proposal, the federal government could seek a court order blocking a model’s release or use when it poses a foreseeable and imminent risk of killing or seriously injuring more than 50 people or causing more than $1 billion in damage. Politico reported that qualifying risks would have to arise from a biological weapon, a cyberattack or a model evading its developer’s controls.
The bill would require frontier-model developers to test their systems, publicly disclose findings, undergo independent evaluations and maintain security programs. Independent evaluators registered with the Commerce Department would periodically assess companies and their models. They would be required to notify the department within 72 hours if they identified an “imminent enumerated risk.”
The Commerce Department could then ask a federal judge in Washington to suspend a model. According to Politico, the draft would prevent other federal departments, agencies or officials from independently halting a model’s development, deployment or internal use.
An Anthropic spokesperson said the legislation reflects proposals in the company’s Advanced AI Framework and would give the government the ability to deter dangerous deployments. The draft would also override some state AI laws, including measures in California. Charlie Bullock of the Institute for Law & AI described the proposal as reasonable but weaker than the competing FRONTIER Act introduced in the House.
HOW THIS STORY WAS MADE
Written from 8 articles, 4 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 402 h after the record first saw the story; the editor kept it as written.
- Publishers
- 128
- Source articles
- 219
- Given to the writer
- 8, 4 with the article's own text
- Independent newsrooms
- 8
- This version written
- 402 h after the record first saw the story
- Second model (editor)
- kept as written
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and published it as written.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →