US Lifts Export Controls on Anthropic's Advanced AI Models After Safety Resolution

The US government has lifted export restrictions on Anthropic's advanced artificial intelligence models, Fable 5 and Mythos 5, just weeks after ordering their suspension over national security concerns. The decision, communicated by the US Department of Commerce, clears the way for wider distribution of these powerful AI tools after Anthropic addressed the government's safety concerns. Access to Fable 5 will be restored globally starting Wednesday, July 1, while Mythos 5 access for US organizations was reinstated on June 26, with plans to expand to international partners.

Export Controls Imposed Over Security Concerns

The restrictions were initially imposed on June 12, 2026, via a private letter from the Commerce Department, requiring Anthropic to obtain US permission before allowing any foreign national, regardless of location, to access the models. In response, Anthropic immediately suspended access to both Fable 5 and Mythos 5 for all users, citing the inability to reliably verify nationality in real-time. The Trump administration's order stemmed from unspecified national security concerns, though Anthropic stated its understanding was that officials were worried about potential vulnerabilities in Fable 5. The company believed the government had become aware of a method of "jailbreaking" Fable 5, a process to bypass software safety restrictions. Anthropic, however, disagreed that a narrow potential jailbreak should warrant recalling a commercial model deployed to hundreds of millions of people. Fable 5 is designed for the consumer market, capable of deep reasoning and performing complex tasks independently. Mythos 5, built on Anthropic's Claude platform, is intended for businesses and cybersecurity experts, with capabilities to identify and exploit vulnerabilities in computer code. These models are rivals to platforms like OpenAI's ChatGPT and Google's Gemini.

Anthropic's Resolution and Enhanced Safeguards

To secure the lifting of the export controls, Anthropic engaged in discussions with US officials and agreed to implement several measures. According to US Commerce Secretary Howard Lutnick, the company committed to proactively detecting and addressing security risks associated with the models, collaborating with the government on standards for upcoming models, and informing authorities of any malicious activity. Anthropic's own statement detailed updates to its safeguards, including the use of safety classifiers to detect potentially dangerous cybersecurity uses of its models. The company emphasized its gratitude to users for their patience and to all who worked on redeploying the models. With the restrictions removed, Fable 5 will be available on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. For paid plans, Fable 5 will be included for up to 50% of weekly usage limits through July 7, then accessible via usage credits. Anthropic is also working to re-enable access on AWS, Google Cloud, and Microsoft Foundry as quickly as possible. Mythos 5 access, initially restored for US organizations operating and defending critical infrastructure, is set to expand to a broader set of domestic and international partners through the Glasswing program.

A Call for Industry-Wide Collaboration and Future Implications

Beyond its immediate resolution, Anthropic is advocating for a shared industry framework to assess and fix potential AI model "jailbreaks." The company argues that a consistent standard for judging the severity of such bypass techniques would help developers triage new findings, launch capable models more safely, and communicate risks transparently to government and industry partners. Anthropic has begun developing this framework in collaboration with Amazon, Microsoft, Google, and other Glasswing partners. Furthermore, Anthropic is strengthening its collaboration with the US government through new pre-release testing, information sharing, and research cooperation. This development for Anthropic's models occurs within a broader context of shifting US export control policies for AI technology. In related news, the US has also lifted sanctions on some AI chip exports to China, including Nvidia’s H20 AI chips and AMD’s MI308 accelerators. While these are less powerful versions of top-end products, the move is expected to significantly boost the US AI chip market, with Nvidia anticipating recouping $15 billion in revenue from H20 chip sales this year. Critics in the US have expressed concerns that easing these controls could diminish the country's technological edge and enhance China's AI capabilities. This episode underscores the intricate balance regulators and AI developers must navigate between rapid innovation and the pressing demands of national security.