Anthropic launched Claude Fable 5 on Monday, June 9, 2026. By Friday, June 12, at 5:21 p.m. Eastern Time, the company received a directive from the U.S. Department of Commerce: block Fable 5 and Mythos 5 for all foreign nationals — both outside the U.S. and within the country, including employees without American citizenship.
What lies behind the "threat to national security"
The government provided no official explanation in the letter. According to reports from Axios and CNBC, the Department of Commerce acted after another company claimed to have found a way to "jailbreak" Mythos, which alarmed the administration about national security risks.
Anthropic conducted its own review and reached a different conclusion. The company argues that this involves a narrow, non-universal jailbreak that reveals Mythos's cyber capabilities only in one specific scenario — essentially asking the model to read certain code and fix vulnerabilities. At the same time, Anthropic confirmed that the same level of capabilities is already available in competing models, including OpenAI's GPT-5.5.
"We believe that the discovery of a narrow potential jailbreak should not be grounds for withdrawing a commercial model deployed to hundreds of millions of people"
— from Anthropic's official statement, June 13, 2026
Why the decision appears selective
OpenAI's GPT-5.5, which according to Anthropic is vulnerable to the same technique, faced no restrictions whatsoever. The administration attempted to convince Anthropic to suspend the release of new models, but without success — and only after that did it resort to the directive.
The context intensifies the questions: Last week, the Financial Times reported that the National Security Agency is already using Mythos to conduct offensive cyber attacks. In other words, the model is simultaneously considered safe enough for NSA operations — and too dangerous for a foreign developer in Anthropic's San Francisco office.
Stakes: IPO and precedent for the entire industry
The timing was no accident from a commercial perspective. Anthropic raised $65 billion in a Series H round in May 2026 at a $965 billion valuation and confidentially filed documents for an IPO expected in October 2026. The directive struck the company's reputation on the eve of its public market debut.
But the broader stakes extend beyond stock prices. For the entire artificial intelligence industry, this establishes a clear precedent: the U.S. government is prepared to treat advanced AI models the same way it treats advanced semiconductors or military technologies — as strategic assets requiring export licensing.
Anthropic emphasized that if such a standard were applied across the entire industry, it would effectively halt the deployment of any new models from all leading developers.
Access to all other Anthropic models remains unchanged. But the question hanging over the entire industry after this directive is: if a jailbreak is discovered in an OpenAI model next time — and the White House remains silent again — will that be evidence of consistent security logic or evidence of selective pressure on a competitor?
