The U.S. Department of Commerce has officially lifted export restrictions on Anthropic's flagship models — Claude Fable 5 and Mythos 5. The decision took effect on June 30, and starting July 1, the company will begin a phased restoration of full global access to these AI systems.
This move is the result of nearly three weeks of intense negotiations between the Anthropic team and White House representatives. As a reminder, the restrictions were imposed on June 12 — just three days after the high-profile release of the models on June 9. Authorities cited national security risks related to potential attempts to bypass AI safeguards. In response, Anthropic took an unprecedented step, completely blocking access to both models for all international users across all its platforms, including Claude.ai, the API, and partner services like AWS Bedrock.
Why was the ban lifted so quickly?
In practice, it turned out that filtering users by nationality in real-time is technically unfeasible without serious operational disruptions. Anthropic chose the only viable strategy — a complete shutdown for the global community until the situation became clear. And this strategy worked. The swift lifting of restrictions demonstrates not only the effectiveness of the company's negotiating position but also the formation of a more mature, dialogue-based approach to regulating advanced AI systems in the U.S.
Official confirmation appeared on the company's social media on June 30:
"We have received notification that the U.S. Department of Commerce has lifted export restrictions on Claude Fable 5 and Mythos 5. Starting tomorrow, we will begin restoring access and will share details soon."
What are Anthropic's flagship models?
Claude Fable 5 is the pinnacle of Anthropic's current lineup. Built on the Mythos-class architecture, the model is equipped with finely tuned safety mechanisms, making it a versatile tool for complex tasks: from big data analysis and programming to image processing. Mythos 5, on the other hand, was initially designed for more sensitive scenarios, particularly in the field of cybersecurity. Access to it was restricted from the start to trusted partners through the Project Glasswing program for critical projects in government agencies and large corporations.
Pricing remains competitive: $10 per 1 million input tokens and $50 per 1 million output tokens. Anthropic's internal classifiers automatically redirect high-risk queries (especially in the areas of bio- and cybersecurity) to safer algorithms across all of the company's products.
My comment as an analyst: This case is an important precedent. It shows that even the strictest regulatory measures can be reconsidered if a company demonstrates a willingness to engage in dialogue and absolute transparency. For the AI market, this is a signal: a balance between innovation and security is possible, but it requires developers to take an active stance rather than passive waiting.