Anthropic's leaked AI bill would weaken safety rules, critics argue
AITransformer argues that a leaked federal AI bill drafted by Anthropic is much weaker than the FRONTIER Act and could make it harder to block dangerous models. The author cites self-certifying evaluators, a court standard that gives no weight to public harm from not blocking a model, and preemption of state laws.
AIWhy it matters.
The piece dissects specific clauses of a leaked federal AI bill, showing how evaluator standards and court-blocking rules shape whether dangerous models can be stopped.
Source details
Transformer · Shakeel Hashim · transformernews.aiPublished · added here
Original title: Anthropic’s weak AI bill undermines its ‘safety first’ approach