It started, as many of 2026's big tech moments have, with a leak.

Back in March, a configuration error in Anthropic's content management system accidentally exposed a draft blog post describing a model the company hadn't yet announced. That leaked document described something called Claude Mythos, a model an Anthropic spokesperson later confirmed marked a "step change" in AI and was the "most capable" system the company had ever built, with what they called "meaningful advances in reasoning, coding, and cybersecurity."

The problem? The cybersecurity part made a lot of people very nervous.

Anthropic was privately warning top government officials that Mythos made large-scale cyberattacks much more likely in 2026. Meanwhile, cybersecurity stocks slumped on the news. The model, by Anthropic's own admission, was that powerful.

A Rough Year with Washington

To understand this week's events, you have to back up a few months. Anthropic and the U.S. government had already been at each other's throats. Earlier this year, Defense Secretary Pete Hegseth issued a directive halting the Pentagon's use of Claude entirely, triggering significant resistance from military personnel and defense contractors who argued Claude outperformed competing systems. The relationship between Anthropic and Washington was, to put it charitably, complicated.

Just days ago, Anthropic launched Claude Fable 5 and Claude Mythos 5 to the public. The excitement was short-lived. On June 12, Anthropic received a letter from the U.S. Department of Commerce signed by Secretary Howard Lutnick. The message was a blunt order: suspend all access to Fable 5 and Mythos 5 for any foreign national, inside or outside the United States, including Anthropic's own foreign national employees. Because Anthropic cannot filter foreign nationals from U.S. users in real time, it had no choice but to disable the models for its entire customer base.

The reason given comes by the government becoming aware of a method of bypassing, or "jailbreaking," the Fable 5 model, and Mythos 5, which has fewer guardrails and is particularly powerful at discovering cybersecurity exploits. This episode marks the first time a U.S. administration has used the export control tool, a mechanism designed for chips and military hardware, against a language model already distributed commercially to hundreds of millions of people.

Anthropic Pushes Back

Anthropic complied, but it didn't stay quiet. The company reviewed a demonstration of the specific jailbreak technique and found it produced a small number of previously known, minor vulnerabilities. These findings appear relatively simple, and other publicly available models are able to discover them without any bypass at all. The company said the government's directive appeared to be based on a narrow, non-universal jailbreak involving capabilities already available in other models, including OpenAI's GPT-5.5.

Anthropic also pointed to the safeguards already in place. The company had required 30-day retention of customer data specifically to research and mitigate jailbreaks, a policy that carries real costs with customers, but one it adopted precisely for this kind of scenario. It stands by its "defense in depth" strategy. The company's core argument is a pointed one: "If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."

The Government's Perspective

David Sacks, co-chair of the President's Council of Advisers on Science and Technology, offered the administration's version of events on X. A highly credible partner testing Fable had come forward with a jailbreak of its guardrails. The administration asked Dario Amodei to fix the jailbreak or pull the model. Amodei refused. Sacks argued that Anthropic's reaction ran counter to the company's own public claims about AI safety, that Anthropic had prioritized keeping a consumer model online over addressing a potential security risk. But Sacks also left a door open, expressing hope that Anthropic would fix the issue and that Fable 5 could return to public release. "The ball," he said, "is in Anthropic's court."

What It All Means

All other Claude models, including Opus 4.8, remain available and unaffected. For everyday users, the disruption is limited.

We now live in a world where a government can issue an emergency directive and pull some of the most capable AI models ever built from the internet within hours. Every company that has woven a frontier AI model into the core of its operations now has a concrete answer to a question they were probably not asking: what happens if the government turns it off. Unfortunately, now the question is when rather than if.

Access can disappear in hours, with no forewarning, no appeal window, and no guaranteed timeline for return. Anthropic built a model so powerful it frightened Washington, and then found itself in a standoff. It's a strange and specific kind of 2026 problem. And it almost certainly won't be the last one.

References:

[1] Anthropic. "Statement on the US Government Directive to Suspend Access to Fable 5 and Mythos 5." Anthropic.com, June 12, 2026. https://www.anthropic.com/news/fable-mythos-access

[2] Al Jazeera Staff and Reuters. "US Orders Anthropic to Disable AI Models for All Foreign Nationals." Al Jazeera, June 13, 2026. https://www.aljazeera.com/news/2026/6/13/us-orders-anthropic-to-disable-ai-models-for-all-foreign-nationals

[3] CNBC. "Anthropic Disables Access to Fable 5 and Mythos 5 to Comply with Government Directive." CNBC, June 12, 2026. https://www.cnbc.com/2026/06/12/anthropic-disables-access-to-fable-5-and-mythos-5-to-comply-with-government-directive.html

[4] Euronews. "What Is Anthropic's Mythos? The Leaked AI Model That Poses Unprecedented Cybersecurity Risks." Euronews Next, March 30, 2026. https://euronews.com/next/2026/03/30/what-is-anthropics-mythos-the-leaked-ai-model-that-poses-unprecedented-cybersecurity-risks

[5] MarkTechPost. "Anthropic Disables Claude Fable 5 and Mythos 5 After US Government Order." MarkTechPost, June 13, 2026. https://www.marktechpost.com/2026/06/13/anthropic-disables-claude-fable-5-and-mythos-5-after-us-government-order

[6] MEXC/CoinCentral. "Pentagon Faces Backlash Over Anthropic Claude AI Ban as Replacement Proves Difficult." MEXC News, March 19, 2026. https://www.mexc.com/news/963405

[7] 9to5Mac. "Anthropic Pulls Claude Mythos 5 and Fable 5 Following US Government Directive." 9to5Mac, June 12, 2026. https://9to5mac.com/2026/06/12/anthropic-pulls-claude-mythos-5-and-fable-5-following-us-government-directive

[8] Pillitteri, Pasquale. "Anthropic Shuts Down Fable 5 and Mythos 5 on US Government Order." pasqualepillitteri.it, June 13, 2026. https://pasqualepillitteri.it/en/news/4838/anthropic-shuts-down-fable-5-mythos-5-us-government-order