US Forces Anthropic to Restrict Best AI Models, Calls It ‘Wake-Up Call for Europe’

In an unprecedented move that has sent shockwaves through the global AI industry, the United States government has forced AI company Anthropic to disable its two most advanced artificial intelligence models — Claude Fable 5 and Mythos 5 — for all users worldwide. The directive, issued by the Commerce Department on June 12, 2026, cites national security concerns over a potential “jailbreak” vulnerability, marking the first time a government has unilaterally recalled a commercial AI product on security grounds.

The Directive and Its Immediate Impact

On the evening of June 12, Anthropic received an export control directive from the Commerce Department ordering it to suspend all access to Fable 5 and Mythos 5 by any foreign national, including foreign national employees of Anthropic itself. Because the company cannot reliably verify user nationality in real time at scale, it was forced to abruptly disable both models for all customers globally — including U.S. citizens and its own workforce.

“That a government dictates who may use AI — there is no precedent for that, to my knowledge,” said Tom Van de Weghe, AI expert at VRT NWS, the Belgian public broadcaster that first reported the story. The only models affected are Fable 5 — launched just three days earlier on June 9 as the first publicly accessible version of the Mythos family — and Mythos 5 for enterprise partners. All other Anthropic models remain operational.

The Alleged Vulnerability

The government’s concern centers on a claimed method to “jailbreak” Fable 5 — bypassing its safety guardrails. Anthropic says it reviewed the technique and found it produced only minor, previously known vulnerabilities that are also discoverable using other publicly available AI models, including OpenAI’s GPT-5.5. The company stated that the government has only provided “verbal evidence of a potential narrow, non-universal jailbreak” and no written technical justification.

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” Anthropic wrote in its official statement. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”

A Troubled Relationship with Washington

The export control directive cannot be understood in isolation. Anthropic’s relationship with the Trump administration had been adversarial for months before this action. In February 2026, a landmark Pentagon deal collapsed after Anthropic refused to allow Claude to be used for lethal autonomous warfare and mass surveillance. Defense Secretary Pete Hegseth then designated Anthropic a “supply chain risk” in March — a label typically reserved for foreign adversaries. Anthropic sued, and a federal judge temporarily blocked the blacklisting.

As Forbes reported, Defense Secretary Hegseth publicly celebrated the ban on June 13, posting: “Three months ago, @DeptofWar kicked @AnthropicAI out of our building — forever. Every passing day proves why that was the right move.”

The Irony of Dario Amodei’s Essay

The timing of events has created a powerful narrative. Just two days before the directive, on June 10, Anthropic CEO Dario Amodei published a policy essay arguing that governments should have the legal authority to block or reverse the release of dangerous AI models, comparing it to FAA aircraft safety standards. The essay called for exactly the kind of government oversight power that was used against Anthropic 48 hours later.

Technology influencer Sara Tortoli captured the market’s reaction: “When you spend years describing your model as potentially civilization-ending, you should not be surprised when governments start treating your model like weapons.”

Amazon’s Controversial Role

Perhaps the most startling revelation is the role of Amazon — Anthropic’s largest investor and cloud partner. According to TechCrunch, the Wall Street Journal, and Reuters, Amazon CEO Andy Jassy raised concerns to Treasury Secretary Scott Bessent and other officials about security risks in Anthropic’s models. Amazon researchers reportedly discovered the jailbreak technique and reported it directly to the Commerce Department rather than to Anthropic through standard coordinated disclosure.

David Sacks, Co-Chair of the President’s Council of Advisers on Science and Technology, confirmed that “a highly credible trusted partner of both Anthropic and the USG” — widely understood to be Amazon — came forward with the jailbreak. Sacks also stated that the administration asked Amodei to fix the vulnerability or de-deploy the model, and that Amodei refused.

A ‘Wake-Up Call for Europe’

The European dimension of this story is particularly significant. Just 12 days before the ban, on June 1, Anthropic had agreed to grant the EU access to Mythos after weeks of negotiations — only for that access to be revoked by a U.S. government directive. The EU had to seek American permission even for this limited access, as CNBC reported.

“For Europe this is a wake-up call,” said Tom Van de Weghe. “Our dependence on the most advanced American models is now painfully visible: one order from Washington can cut off access to the best systems overnight. That directly affects our competitive position and our digital autonomy.”

The European Commission had adopted its ambitious Tech Sovereignty Package just nine days before the export control, on June 3, including the Chips Act 2.0 and the Cloud and AI Development Act (CADA), aimed at reducing Europe’s reliance on non-EU digital technology. The Anthropic ban gives this initiative new urgency.

Implications for AI Governance

The precedent set by this action is profound. Previous U.S. export controls on AI focused on hardware (chips to China) and model weights. This directive targets a deployed commercial service — effectively a product recall based on a verbally described concern about a narrow jailbreak.

As explainx.ai noted, Anthropic’s detailed safety documentation — including honest admissions that perfect jailbreak resistance is impossible — may have supplied the evidence for the government’s action. This creates a perverse incentive: companies that are less transparent about their models’ limitations face less regulatory risk. If this becomes the industry’s lesson, it could harm AI safety transparency across the board.

What Comes Next

Anthropic says it believes the situation is “a misunderstanding” and is working to restore access. David Sacks has indicated that the resolution path is straightforward: Anthropic patches the jailbreak, the export control is lifted, and Fable 5 returns. But several critical questions remain unanswered.

Will the same standard be applied to OpenAI’s GPT-5.5, which Anthropic claims can perform the same tasks? Will this accelerate investment in European AI alternatives like Mistral AI and open-source models? And how will the existing legal battle between Anthropic and the Pentagon intersect with this new action?

For now, the global AI industry faces a new reality: the most advanced models can be switched off by government decree, and dependence on American AI technology carries geopolitical risk that no contract can mitigate.