The world’s most advanced public AI was just yanked offline overnight by Washington lawyers waving a national security letter.

Story Snapshot

• The Commerce Department ordered Anthropic to block foreign nationals from using its top models, Fable 5 and Mythos 5.
• Anthropic says the order arrived at 5:21 p.m. Eastern and left them no choice but to shut the models down for everyone.^[3]
• The government’s concern is a narrow “jailbreak” that Anthropic says is minor and already possible in other public AIs.^[2]
• This marks a new phase where export controls built for weapons now reach into cloud software and everyday AI tools.

How Washington Reached Into the Cloud and Turned Fable 5 Off

On Friday evening, Anthropic, the company behind the powerful Fable 5 and Mythos 5 models, posted a stunning notice: the United States government had issued an export control directive ordering them to suspend all access to both models by any foreign national, inside or outside the country.^[2] The letter arrived at 5:21 p.m. Eastern, and the company said the only way to comply fast was to shut the models off for everyone, worldwide.^[3][4] Access to Anthropic’s other models stayed live.^[2]

Commerce Secretary Howard Lutnick’s letter reportedly puts Fable 5 and Mythos 5 under the same export license rules long used for advanced weapons and chips, but now applied to cloud software.^[3] The order covers foreign nationals even when they are physically inside the United States and even Anthropic’s own non‑citizen employees, turning normal product access into a legal minefield.^[4] Overnight, the world’s most capable public models went dark not because of a technical failure, but because of Washington paperwork.^[3]

What “Jailbreak” Triggered the Shutdown – and Why Anthropic Disagrees

Officials say the directive was sparked by a report from another company claiming it could “jailbreak” Mythos 5, using the model to spot software flaws that could become cyber weapons in the wrong hands.^[3] Anthropic confirms the government believes it learned of a method to bypass Fable 5’s safeguards but says the demo only surfaced a few minor vulnerabilities that were already known and easy to find.^[2] The company adds that other public models like OpenAI’s latest systems can do the same thing today.^[4]

Anthropic stresses that no one has shown a “universal jailbreak” that blows past the model’s protections across the board, and it has not seen a single case where a non‑universal jailbreak led to real‑world harm.^[2][4] In its statement, the company flatly calls the government’s move a “misunderstanding” and warns that if this standard is applied across the industry, it would freeze deployment of nearly all new top‑tier AI models.^[2][5] For everyday users and businesses, that means access to safer, more capable tools can vanish, even when the technical risk looks narrow and manageable.

Export Controls: From Chips and Missiles to Lines of Code

For decades, export controls focused on physical items like high‑end chips, missiles, and factory tools, with rules written under the Export Administration Regulations and enforced by the Commerce Department’s Bureau of Industry and Security. Since 2022, those rules have tightened sharply on advanced semiconductors, especially where China is involved, adding broad “catch‑all” bans when officials think a technology could support foreign militaries or intelligence services. Legal experts now describe a clear trend: Washington is willing to stretch old tools to cover new threats in artificial intelligence.

Policy analysts say the Fable 5 order fits a pattern of ad hoc, model‑specific directives aimed at so‑called “frontier” systems that might help build cyber weapons or advanced surveillance tools. The Remote Access Security Act, passed by the House in 2026, goes even further by explicitly extending export control reach to remote access over the internet or cloud when foreign users pose a security risk. For conservative readers, that should raise alarms: the same bureaucracy that once licensed tank parts is now deciding, in secret letters, who may talk to a math engine in the cloud.

What This Means for Innovation, Liberty, and the Next AI Fight

Anthropic’s shutdown came just three days after launching Fable 5 to the public, blindsiding developers, researchers, and companies that had begun testing the model for everything from coding help to security defense.^[3] Commentators across the tech world warn this is the most far‑reaching action any government has taken against a commercial AI model and could become the template for future crackdowns whenever officials feel uneasy. Some warn that if every new model risks a last‑minute export squeeze, companies will hesitate to ship their best work or will limit it to a small, approved group of users.^[5]

On Friday night, the tech world saw something completely unprecedented: the global, sudden shutdown of Anthropic’s Claude Fable 5 and Mythos 5.

Not throttled. Not fenced. Switched off globally by an undisclosed government export-control directive—just three days after launch.…

— ARAGS Inc. (@ARAGS_Inc) June 15, 2026

Supporters of strict controls argue that frontier AI can already help find software holes, design biological threats, or guide cyber attacks, and say fast, decisive action is needed to keep those tools away from hostile states. But even they rely on clear rules and due process, not vague letters with little evidence attached. Anthropic says it is obeying the law while fighting to restore access and pressing the government for more detail.^[2][3] For citizens who care about free inquiry and limited government, the core question now is simple: who decides when your tools are “too smart” for you to use—engineers you can hold accountable, or unelected regulators acting in the dark?

Sources:

[2] Web – Statement on the US government directive to suspend access to …

[3] X – Anthropic

[4] Web – Anthropic Fable 5 Shutdown: US Export Order Forces a Global …

[5] Web – US Export-Control Order and Global Suspension of Fable 5 …