US lets Anthropic re-release Mythos 5
The US government cleared Anthropic to give its strongest model, Claude Mythos 5, to a vetted group of trusted partners, partly reversing an early-June block.
On June 26, the US government cleared Anthropic to release Claude Mythos 5, its strongest cybersecurity model, to a vetted set of partners. It partly reverses an export-control order from earlier in June that made Anthropic disable Mythos 5 and Fable 5 for everyone over national-security concerns.
In a letter, Commerce Secretary Howard Lutnick wrote that “appropriate safeguards are in place” to let certain trusted partners use the model. The cleared group is described as cyber defenders and critical-infrastructure providers; reporting put it at more than 100 organizations, many of them Fortune 500 companies. Fable 5, described as a less powerful version, was not included, and talks over restoring broader access were said to be continuing.
What it means for you
This is the part operators keep underrating: the ceiling on what you can build is now partly a policy question. If the most capable model ships only to a vetting list, your access can change without your product changing at all.
It is the same dynamic behind export controls on top Claude models and the White House asking OpenAI to limit GPT-5.6. The durable lesson for a working operator is to plan for the day your model gets switched off.