Trump forces Anthropic to take AI models offline

What happened

The Trump administration sent an export control directive to Anthropic, forcing the company to pull its Claude Mythos and Fable 5 models offline. After days of negotiations, Anthropic and the White House remain at odds over how to bring them back. The White House demanded that Anthropic prohibit all foreign nationals from accessing these models, locking out employees, customers including Apple and Meta, and preventing further research use.

Why it matters

The US government has not clearly stated what Anthropic did wrong—the dispute is defined by its opaqueness. The core issue appears to be that Anthropic shared Mythos with SK Telecom, a South Korean telecom giant, and that some safeguards on Fable 5 could allegedly be circumvented. However, Anthropic says it coordinated the SK Telecom rollout with the government and revoked access immediately when concerns were raised. The real problem is that the Trump administration, which has opposed AI regulation, now finds itself making regulatory decisions in real time without established procedures, putting the entire AI industry in a quandary about what rules actually apply.

What to watch

Other major AI labs like OpenAI, Google, and Meta have been watching closely and are reaching the same conclusion: they believe they will need to give the White House early access to their latest models and be proactive about sharing information about upcoming launches. President Trump signed an executive order last month creating a 'voluntary' system for AI labs to submit models for early government testing, but Anthropic's experience suggests this has effectively become a mandatory licensing regime.