The U.S. Department of Commerce has officially approved the lifting of export restrictions on Anthropic's flagship artificial intelligence models — Claude Fable 5 and Mythos 5. The decision took effect on June 30, and full global access to these systems will be restored starting July 1.
This development resulted from nearly three weeks of intensive negotiations between the Anthropic team and White House representatives. As a reminder, the restrictions were imposed on June 12, just three days after the models' release on June 9. Authorities cited national security risks due to potential attempts to bypass AI safeguards.
What's behind the decision?
Export restrictions are a standard U.S. practice for controlling the spread of sensitive technologies. However, the situation with Claude Fable 5 and Mythos 5 proved unique. The imposed ban almost immediately caused operational disruptions: it turned out that filtering users by nationality in real time was technically impossible. As a result, Anthropic was forced to completely disable both models for all clients worldwide — via Claude.ai, API, AWS Bedrock, and other partner services.
Official confirmation from the company was not long in coming:
"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," Anthropic stated on June 30.
Claude Fable 5 and Mythos 5: In the spotlight
Claude Fable 5 is Anthropic's most advanced model, built on the Mythos-class architecture with additional finely tuned safety mechanisms. It demonstrates outstanding results in complex tasks, handling large volumes of data, programming, and image analysis.
Mythos 5, in turn, is designed for more sensitive scenarios. Built on the same architecture, it adds extra safety layers, making the model suitable for cybersecurity. Access to it was initially restricted to trusted partners through Project Glasswing for critical projects in government agencies and large corporations.
Pricing for both models remains competitive: $10 per million input tokens and $50 per million output tokens. Internal classifiers automatically redirect high-risk requests to safer algorithms — especially when processing tasks related to cybersecurity and biology.
My analysis: This decision is an important signal for the entire industry. The swift lifting of restrictions demonstrates not only the effectiveness of Anthropic's negotiating position but also the formation of a more mature approach to regulating advanced AI systems in the U.S. The market gains access to cutting-edge models without excessive bureaucratic barriers, which will undoubtedly accelerate innovation in the field of artificial intelligence.