The White House Is Making Up Its Rules for AI in Real Time
TL;DR
Anthropic still cannot distribute Claude Mythos or Fable 5 after a White House export-control directive; WIRED says no concrete rule violation has been named. The dispute centers on SK Telecom access, alleged China links, and claims that Fable 5 guardrails can be bypassed through jailbreaks. The administration reportedly told Anthropic to block all foreign nationals from the models, including some of its own employees.
Nauti's Take
Export control by improvisation is poison for any frontier roadmap. If Washington freezes models first and explains later, compliance becomes a product feature.
Build launches as if a regulator may ask tomorrow for admin access, user lists, and jailbreak demos.
Briefingshow
The US has tried to steer frontier AI through voluntary testing and pro-innovation language. The Anthropic case shows the harder reality: when formal rules are thin, political officials decide case by case what safety and export limits mean. For labs, customers, and international teams, that can be harder to plan around than strict but clear regulation.
Sources