Anthropic pitches AI safety bill that may block a model if 50+ people could be killed

Anthropic has proposed an AI safety bill that would require federal agencies to block a model's release if it could potentially cause harm to 50 or more people, but with a catch: the agency would need to get a judge's permission to do so. This effectively means that the protection against AI run amok may not be as robust as it seems, as a judge could potentially override the decision.