Leaked Anthropic AI bill: action only at $1B or 50 harmed

AI safety rules are moving from theory to thresholds. A leaked draft of an AI safety bill from Anthropic reportedly ties government intervention to situations involving 50 or more people harmed, or $1 billion at risk. For anyone working in AI, policy, or product, that's the detail to watch. Thresholds like these could shape when companies escalate risks, how testing is scoped, and where compliance teams focus their resources. Vague standards create uncertainty, while specific numbers create clearer obligations but also invite debate over where the line belongs. The story is based on a leak, so the wording could still change. Still, it shows how one major AI developer is framing the moment when outside oversight should begin. Where do you think that line belongs?