The US government has cleared Anthropic to restore limited access to its powerful Mythos 5 AI model after the company addressed national security concerns over potential jailbreaking. The clearance, detailed in a letter from Commerce Secretary Howard Lutnick, ends a two-week standoff that began when the government abruptly barred foreign nationals from using Mythos 5 and its sibling model, Fable 5. Anthropic had disabled global access to both systems and entered intensive negotiations to resolve the dispute. The Commerce letter does not mention any change to restrictions on Fable 5, which remains unavailable for general use.
The resolution allows Mythos 5 to be redeployed to a small group of cyber defenders and infrastructure providers under a program called Project Glasswing. About 200 firms, including Apple, Google, and Microsoft, had previously accessed the model. Anthropic is working to restore access quickly and continues discussions with the government to expand availability and bring back Fable 5. The company is also collaborating on a policy framework to handle similar situations in the future, though it remains unclear what specific steps Anthropic took to satisfy security concerns.
The government's intervention marks the most significant US action to date into an AI venture's operations, highlighting emerging security fears around cutting-edge models. The dispute reverberated across the industry, with rival OpenAI also limiting release of its GPT-5.6 model under administration pressure. Anthropic, which has long positioned itself as a responsible AI developer, faces ongoing friction with the Pentagon over military use of its tools. The company recently filed confidentially for an IPO at a valuation topping $900 billion, making the regulatory outcome critical to its market positioning.
What to watch next: Whether the government lifts restrictions on Fable 5 and how the emerging policy framework shapes future AI model deployments across the industry.
Key Takeaways
- Anthropic regained limited access to Mythos 5 after two weeks of negotiations with the Trump administration over national security risks.
- The government's intervention represents the most significant US regulatory action to date against an AI company's operations.
- Fable 5 remains restricted, and Anthropic is working to bring it back while developing a future policy framework with the government.
- The dispute pressured rival OpenAI to limit its GPT-5.6 release, signaling broader industry impact from heightened AI security scrutiny.
Insights & Analysis
- The government's willingness to impose export controls on frontier AI models signals a new era of proactive regulation, potentially slowing the pace of commercial AI releases and reshaping competitive dynamics between US and foreign AI firms.
- Anthropic's ability to negotiate a partial resolution while maintaining its IPO trajectory suggests that regulatory risk, while significant, can be managed through direct engagement—but the lack of a clear standard for 'jailbreak resistance' leaves all model providers vulnerable to future shutdowns.