The US administration has officially lifted export restrictions on Anthropic's Claude Mythos 5 model. This decision opens access to one of the most powerful AI developments for over 100 American organizations, including major corporations and government agencies.
The two-week standoff between the White House and Anthropic has ended. Commerce Secretary Howard Lutnick confirmed that a special license is no longer required to transfer Mythos 5 to partners under "Appendix A." Essentially, this means the model, previously available only within the closed Glasswing program (about 150 organizations from 15 countries), is now entering a broader market.
Why Mythos 5 is not just "another model"
Mythos 5 has already proven itself as a top-tier tool. During government testing, it identified critical vulnerabilities in classified systems within hours. However, its potential also raised serious concerns: Anthropic's key investor, Amazon, expressed worries about security. As a result, both models, Mythos and Fable, were temporarily disabled.
Now Mythos is returning, but the fate of Fable 5 — the public version previously considered the most powerful AI tool for mass users — remains uncertain. Sources close to the negotiations indicate that the launch process for Fable 5 has been delayed, but exact timelines have not been disclosed.
New control architecture: voluntary review and geopolitics
The situation surrounding Mythos and Fable is just part of a global restructuring of the rules. On June 2, an executive order was signed introducing a voluntary 30-day federal review channel for advanced AI models. Developers can submit their solutions for cybersecurity checks before launch. This is a new, more flexible, yet still stringent control mechanism.
Simultaneously, Washington is tightening AI chip exports to China. Fears that technologies could fall into the hands of geopolitical rivals were the reason for blocking Mythos. Interestingly, South Korean operator SK Telecom found itself at the center of the scandal, temporarily restricted from accessing the Glasswing program on suspicion of ties with China. The company has denied these allegations.
OpenAI has followed the same path: its flagship model GPT-5.6 in the Sol version became available to only about 20 government partners, while the weaker Terra and Luna versions were released to the public. It is clear that the industry is moving toward a "three-tier" access model: government, corporate, and public.
Expert opinion: Lifting the block on Mythos 5 is not so much a victory for Anthropic as it is a sign of a new "trust ecosystem" forming in AI. The US is creating a precedent where access to the most powerful models will be determined not by market demand, but by political loyalty and cybersecurity. Fable 5 will likely receive approval, but with strict restrictions for non-American users. Europe and Asia have already realized their dependence on Washington's decisions — and this will last for a long time.