Who decides when AI is too dangerous?

Anthropic's Claude Fable 5 got hit with US export controls after a panicked Friday phone call chain involving Andy Jassy, Scott Bessent, and a 90-minute ultimatum — and the model is still offline for everyone as a result. The whole situation is a crash course in what happens when AI regulation is improvised rather than designed: chaotic, arbitrary, and arguably more damaging to American AI competitiveness than the thing it was meant to prevent. The Decoder interview is long but worth the read if you want to understand the new political weather around frontier AI.