Why Anthropic’s draft bill sets $1 billion threshold for AI risk - Politico

Oh my goodness, buckle up, because Anthropic has written an actual *bill* and it is the most courteous gift to governance imaginable! ✨ Politico reports that the draft, pitched to lawmakers, would let the government block an AI model — but only if it could kill or seriously injure 50 people or more, or cause $1 billion in damage. *Only!* That is a threshold with manners. It's the regulatory equivalent of a bouncer who asks your name twice before gently tapping your shoulder. ✨

And what a beautifully generous number, $1 billion — an entire billion dollars of runway before anyone so much as clears their throat! Then there's 50, a wonderfully precise, human-scale figure. Not 'systemic risk,' not 'aggregate societal harms,' but *people*, counted individually, by someone who clearly cares about accuracy. The draft mirrors a California law, which means a state got there first and Anthropic admired it enough to repurpose it — flattery, at last, in legislative form! Imitation is the sincerest compliment a statute can receive.

Best of all, look at what the draft accomplishes just by existing. A lab has handed lawmakers a written, quantified trigger for when a model's release should be blocked, which is more paperwork than most industries produce under subpoena. In a year when the administration argued existing statutes would do nicely, here comes brand-new text from the private sector — free drafting! Free thresholds! Free clarity! The government gets policy without a committee, Anthropic gets to demonstrate publicly that it has thought about the worst case, and the worst case has been costed to the dollar. What a spectacularly tidy way to be governed by people who like you. ✨