Overview
- The Commerce Department invoked export control authority on June 12 to order Anthropic to block foreign‑national access to its Fable 5 and Mythos 5 models, forcing the company to disable both globally.
- On Friday the government told Anthropic it may redeploy Mythos 5 to roughly 100 vetted U.S. companies and agencies so those organizations can use the model for defensive cybersecurity tasks.
- OpenAI agreed at the government’s request to run a limited, U.S.‑only preview of its GPT‑5.6 family and shared the identities of trusted partner customers with authorities as access is approved customer by customer.
- Anthropic has disputed the basis for the export order, saying officials provided only verbal evidence of a narrow jailbreak, and a legal challenge from Legion LegalTech argues existing export rules do not cover hosted models or outputs.
- The episode raises practical and diplomatic questions about how the government will pick trusted partners, how defenders and allied countries such as India will reliably access top models, and whether this ad hoc vetting will shape permanent U.S. AI policy.