On Friday, the U.S. Department of Commerce officially lifted the two-week ban on the export of Anthropic's Claude Mythos 5 model. Now, access to this advanced AI system will be granted to over 100 organizations, including major corporations and government entities that are part of the so-called "List A."
Commerce Secretary Howard Lutnick sent a letter to Anthropic's Chief Technology Officer Tom Brown, explaining that current precautionary measures allow trusted partners to work with Mythos 5 without additional licensing. This decision ended the two-week standoff between the Trump administration and the company.
Mythos 5: From Glasswing to Broad Access
Recall that Mythos 5 was previously used exclusively within the Glasswing program—a vulnerability discovery initiative involving about 150 organizations from more than 15 countries. Within just a few hours of testing, this model identified critical errors in classified government systems, which raised concerns among regulators.
Now, restrictions have been lifted, but only for American partners. The situation with Fable 5—a more powerful public version—remains uncertain. Unlike Mythos, Fable 5 was previously available to all subscribers and was long considered the most powerful publicly accessible AI tool. However, after Amazon, one of Anthropic's key investors, expressed security concerns, both models were taken offline.
New Rules of the Game: Voluntary Review and Control
On June 2, an executive order was signed that opens a voluntary channel for federal review of advanced AI models. Developers can submit their solutions for cybersecurity checks 30 days before launch. This is a soft but telling step: Washington is clearly tightening control not only over the export of AI chips to China but also over access to the models themselves.
OpenAI followed the same path on Friday: the most powerful version of GPT-5.6, named Sol, became available only to about 20 partners who received state approval. The weaker versions, Terra and Luna, are available to the general public.
Notably, the reason for the blocks was concern that competitors could gain access to the technology, including through South Korean operator SK Telecom, which was temporarily added to the Glasswing list and then had its access restricted. SK Telecom denies any connection to China.
Leaders of major cybersecurity companies, including Nvidia, Adobe, and Zoom, signed an open letter calling for the removal of restrictive measures. However, European countries and other regions are already feeling the dependence on Washington's decisions.
My analysis: Lifting the ban on Mythos 5 is a compromise that does not solve the main problem. Fable 5, as a more powerful and accessible model, remains in question. The AI market is clearly moving toward segmentation: ultra-powerful systems for the select few, weakened versions for everyone. This creates a new class of "digital inequality," where access to advanced technologies becomes a privilege, not a right.