The U.S. government has lifted export restrictions on Anthropic's latest AI model, Claude Fable 5. As a result, global services, which had been suspended for about three weeks, have resumed.
On June 12, the U.S. Department of Commerce imposed export controls to prevent the provision of Claude Fable 5 and Claude Mythos 5 to overseas users for national security reasons. At the time, Anthropic temporarily suspended services for both models worldwide, citing the difficulty of distinguishing users' nationalities in real-time.
With this measure, Claude Fable 5 is now available again on Claude.ai, Claude Code, Claude Platform, and Claude Cowork. Deployment via AWS, Google Cloud, and Microsoft Foundry will also resume sequentially. However, the more secure Claude Mythos 5 will be provided on a limited basis, primarily to select organizations approved by the U.S. government.
Along with the resumption of services, Anthropic has also implemented new safeguards. It added a classifier that detects attempts to find software vulnerabilities or generate attack code using specific prompts, and modified the system to automatically switch dangerous requests to other, lower-performance models. The company explained that this method blocks more than 991 TP3T of known evasion attacks.
This incident is considered one of the first instances where the U.S. government directly controlled a cutting-edge AI model from a national security perspective. Moving forward, it is highly likely that government safety verification and regulatory compliance will become critical factors alongside performance competition when launching new AI models. While users can now use Claude Fable 5 again, it appears that some features will be provided under stricter safety policies than before.
More and Sources
