US Government Lifts Restrictions on Anthropic's Claude Fable 5 and Mythos 5 AI Models

Anthropic, a leading artificial intelligence company, confirmed on Tuesday that the U.S. government has rescinded the export controls previously placed on its highly capable Claude Fable 5 and Mythos 5 AI models. This move brings an end to a contentious period that saw the models temporarily pulled from public access. The company stated that it would commence restoring access to Fable 5 and Mythos 5 starting Wednesday, July 1st, making them available across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork. The dispute originated approximately two and a half weeks prior when the U.S. Commerce Department issued an order requiring Anthropic to restrict access to these models for all foreign nationals, both inside and outside the United States. This broad directive, which even impacted Anthropic's non-citizen staff, led the company to disable the models entirely, as real-time nationality verification of every user proved impractical. The government's concerns were primarily centered on national security risks, specifically the potential for "jailbreaks" – prompts designed to bypass the models' safety protocols. An Amazon research team reportedly discovered such a vulnerability in Fable 5, demonstrating its ability to identify software flaws and, in one instance, generate code illustrating how these flaws could be exploited. While Anthropic downplayed the severity, arguing similar requests work on other models like Claude Opus 4.8 and OpenAI's GPT-5.5, the government deemed the finding serious enough to warrant emergency controls. To address these concerns and facilitate the lifting of restrictions, Anthropic developed and implemented a new safety filter, or classifier. This filter is specifically designed to detect and block the exact jailbreaking technique identified in the report, achieving over 99% effectiveness as of June 30th. Requests flagged by this filter are now routed to the less powerful Opus 4.8 model, and users are notified. While this solution may lead to more false positives in normal coding and debugging scenarios, it satisfied the government's requirements. Mythos 5, which features fewer safety guardrails, will remain under tighter access controls, initially available to a select group of U.S. companies and federal agencies.