Trump Orders Global Shutdown of Anthropic's AI Models: Historic Precedent in Export Controls

What happened?

On June 12, 2026, the U.S. Department of Commerce issued an unprecedented order: suspend global access to Anthropic's Fable 5 and Mythos 5 models, just three days after the commercial launch of the former. The order, signed by Commerce Secretary Howard Lutnick, invokes national security authorities and prohibits any foreign national from accessing these models, even from within the U.S. To comply, Anthropic opted for a total blackout: it deactivated both models for all customers worldwide, unable to verify each user's nationality in real time.

Why did the government intervene?

According to sources cited by Axios, the decision was based on an alleged jailbreak reported by another company, which supposedly managed to bypass Fable 5's security mechanisms. This model, designed for autonomous video game generation, is built on Mythos 5 — Anthropic's frontier model, restricted to vetted organizations through Project Glasswing. The government fears the jailbreak could allow access to Mythos 5's offensive cybersecurity capabilities via Fable 5. Anthropic disputes the severity of the jailbreak and claims its team reviewed the demonstration without finding conclusive evidence.

Why is this important?

This event is historic: it is the first time a Western government has applied real-time export control on a commercial AI already deployed. Until now, controls were applied before launch or through export licenses. Here, access to an active product is revoked, setting a precedent the industry feared. It implies that any frontier AI model can be globally disconnected if the government deems it a national security risk, creating regulatory and business uncertainty.

Immediate consequences

• For Anthropic: loss of revenue from two flagship models, reputational damage, and potential impact on investor and client relationships. The company must renegotiate conditions with the government to reactivate the models.
• For the industry: other companies like OpenAI, Google DeepMind, and Meta will need to review their security mechanisms and contingency plans for potential similar orders. The debate on frontier AI regulation accelerates.
• For users: developers and companies relying on Fable 5 for video games or Mythos 5 for advanced research lose access, potentially delaying projects.
• For the global market: the notion that frontier AI is subject to geopolitical controls is reinforced, potentially fragmenting the tech ecosystem between blocs (U.S., China, EU).

Context and comparisons

This event recalls the 2022 ban on exporting NVIDIA chips to China, but goes much further: it's not about hardware, but software and models already in production. It also resembles the shutdown of TikTok in India in 2020, but with the difference that here the order comes from the U.S. government itself against a domestic company. The AI community has been warning for months that export controls could extend to closed-source models; this case confirms those fears.

What readers should know

The order does not affect Anthropic's other models (like Claude 4 or earlier versions). The government has not presented public evidence of the jailbreak, and Anthropic questions it. The final decision may be resolved in court or through negotiations. Meanwhile, any company developing frontier AI should prepare for potential similar interventions, and users should diversify their AI providers to mitigate risks.

"This is the kind of precedent the industry had been fearing for months, and it has arrived." — TheVortiq