The Trump administration has ordered Anthropic to pull its two most advanced AI models offline, barring all foreign nationals from accessing Fable 5 and Mythos 5 — effective immediately. The move marks one of the most aggressive uses of U.S. export control powers against a domestic AI company to date.

TL;DR: Commerce Secretary Howard Lutnick issued a directive Friday ordering Anthropic to disable Fable 5 and Mythos 5 for all foreign nationals. Anthropic complied within hours. The trigger: a report that a third-party company had successfully jailbroken the Mythos model.

What Happened

Anthropic CEO Dario Amodei received a letter from Commerce Secretary Howard Lutnick on Friday afternoon. The directive, delivered at 5:21 p.m. ET, instructed Anthropic to immediately suspend all access to Fable 5 and Mythos 5 for any foreign national — whether outside the United States or working inside the country.

Anthropic announced compliance within hours, calling it an "abrupt" but legally binding order. The company said it had to disable access for all customers globally to ensure full compliance, given the difficulty of real-time nationality verification at scale.

Why the Government Acted

The trigger was a report from a separate, unnamed company claiming to have successfully jailbroken Anthropic's Mythos 5 model. Administration officials expressed alarm that the model's capabilities — once unlocked from its safety guardrails — could pose a national security risk if accessible to foreign actors.

The Commerce Department invoked export control authority that has historically been applied to physical hardware and semiconductor technology, now extending it explicitly to AI model weights and API access. Legal experts noted this is among the first times a U.S. AI model has been restricted under export control law in real-time, rather than through slower regulatory rulemaking.

The Bigger Dispute

This ban doesn't exist in a vacuum. It escalates a months-long conflict between Anthropic and the Trump administration.

The friction began when Anthropic publicly declined requests to allow U.S. military agencies to use its models for fully autonomous weapons systems — a stance the company said was consistent with its responsible AI principles. The Pentagon responded by placing Anthropic on a government blacklist, classifying the company as too high-risk for government procurement.

Friday's export control order is seen by observers as a continuation of that dispute, using regulatory tools to pressure a company the administration views as uncooperative.

Who Is Affected

Fable 5 and Mythos 5 are Anthropic's most powerful publicly available models as of mid-2026. They power a significant portion of Anthropic's enterprise and developer revenue through Claude.ai and the API. Taking them offline for any foreign national effectively cuts off access for users and businesses across Europe, Asia, Latin America, and beyond.

Anthropic did not announce a timeline for restoration. The company said it is "actively working with legal counsel and the Commerce Department to understand the scope of the directive and seek a path toward compliance that minimises disruption."

What This Means for AI Access

The order sets a precedent that should concern the AI industry broadly. If the U.S. government can compel a domestic AI lab to pull its flagship models from global users within hours, the entire model of globally-accessible AI-as-a-service is now subject to executive branch override.

For Anthropic specifically, this is both a financial and reputational crisis. Its enterprise clients outside the U.S. are now locked out of tools they may have built core workflows around. Competitors haven't commented publicly, but the order has implications for OpenAI, Google DeepMind, and Meta, all of which operate globally available AI models.

Key Takeaways

  • The Trump administration ordered Anthropic to immediately disable Fable 5 and Mythos 5 for all foreign nationals
  • The trigger was a claimed successful jailbreak of the Mythos 5 model by a third-party company
  • Anthropic complied within hours of receiving the directive at 5:21 p.m. ET Friday
  • This escalates an existing dispute over Anthropic's refusal to support autonomous weapons
  • Among the first uses of U.S. export controls to restrict real-time access to an AI model API
  • No restoration timeline has been given

Conclusion

Anthropic now finds itself at the centre of the most direct collision yet between AI development and U.S. national security policy. The practical outcome is clear: the era of frictionless global access to frontier AI models may be over, and the terms under which that access exists are now subject to executive decision-making at a moment's notice.