Anthropic to Revise Leaked AI Safety Bill After Backlash

Anthropic will revise parts of its draft federal AI safety bill circulated among US lawmakers after criticism from AI safety advocates, according to The Next Web. Federal affairs lead Anthony Cimino said the company will clarify language that could override state AI safety laws and re-evaluate its risk thresholds, which were taken from California law.

The draft would allow the Commerce Department to ask a federal court to block a model only when it poses a foreseeable and imminent threat involving more than $1 billion in damage, or death or serious injury to more than 50 people. Covered threats would need to involve biological weapons, loss of developer control or cyberattacks.

Critics also objected to provisions allowing independent evaluators to certify themselves and receive payment from the labs they assess. Cimino said Anthropic plans further proposals on evaluator requirements and measures to prevent companies from selecting less demanding evaluators.