The US Government Shuts Down Anthropic's AI Tools for Non-US Users

The Trump administration on Friday, June 13, 2026, ordered Anthropic to block access to its Mythos AI platform and the consumer-facing Fable 5 chatbot for anyone outside the United States. The government invoked export control authority — a legal power normally used to restrict military or sensitive technology — in what amounts to one of the most direct government controls ever placed on a commercial AI product.

The order came from White House concerns, reported by Semafor, that Chinese nationals or companies might gain access to Mythos. The suspension covers both the enterprise-grade version (Mythos, used by businesses) and its consumer version (Fable 5, used by individuals). This dual scope suggests the government views both tools as potential pathways for sensitive technology to leave the country, not just the most advanced research systems.

Trump adviser David Sacks said publicly that the restrictions are not connected to earlier policy concerns about AI exports, though he did not explain what specific threat drove the June 13 action.

The gap between what Sacks said and what the White House had previously stated about AI export risks is worth noting. If the administration's concerns are genuinely new, it has not yet offered a clear explanation of what the actual threat is. This matters because the legal justification for the order will determine how much disruption Anthropic faces and whether other US AI companies will face similar orders.

How Export Controls on AI Actually Work

The US government restricts access to certain dual-use technologies — things that can have both civilian and military applications — through something called the Export Administration Regulations. The Commerce Department runs this system. In the final days of the Biden administration, officials tightened these rules specifically for frontier AI models (the most capable systems from companies like Anthropic and OpenAI). The Trump administration has revised those rules further, making the restrictions even stricter.

What is unusual here is applying those restrictions to a consumer product like Fable 5 based purely on a user's nationality. The government would need to verify that every user is actually a US citizen — not just checking their country from their internet address, but confirming identity — when they sign up. For a free or low-cost consumer product with millions of potential users, that is a genuine engineering and compliance headache. Most platforms use IP address geofencing (blocking access based on where the user is located) as a first line of defense. Citizenship verification is far more intrusive and harder to do reliably at scale.

What Happens Next

Anthropic finds itself in a tight spot. The company has taken money from US government AI funding programs while also trying to build products for users around the world. This tension is now out in the open.

Competitors like OpenAI, Google DeepMind, and Meta will read this order as a signal about what the administration expects. If this legal approach gets applied to them too, international users of American-made AI tools could lose access with little warning and no easy remedy.

The impact on Anthropic's business customers outside the US will be immediate and painful. Mythos is embedded in workflows at organizations in Europe, Asia, and elsewhere. A sudden shutdown — rather than a phase-out period — would disrupt real work at real companies.

There is a useful pattern to step back and observe here. Over the past few years, the US government has restricted sales of the most advanced computer chips to China, particularly the kind used to train large AI models. This created friction with US allies, but Washington has been willing to accept that friction to slow down Chinese AI development. Now the government is extending the same thinking to the AI models themselves, not just the hardware. It is a logical next step that most people in the industry did not expect to move this quickly.

Two questions will shape what June 13 actually means in practice. First, will Anthropic successfully defend the order in court? Second, can the company figure out how to verify that users are US citizens at scale without losing huge numbers of customers in other countries? Both answers will depend on technical and legal factors that are still unfolding.