Washington Just Locked Europeans Out of Its Best AI. By Decree.

Three days ago, Anthropic released Claude Fable 5 — its most capable AI model ever, available to the public for the first time. Three days ago, European companies, European researchers, European developers and European citizens could access frontier American AI on equal terms with their American counterparts.

That ended on Friday evening, June 12, 2026.

What Happened

Anthropic was forced to disable all access to its newest AI models, Fable 5 and Mythos 5, late on Friday after the US Commerce Department used national security export controls to bar the company from distributing the models to any foreign national. The directive includes not just people located outside the US, but also any foreign national in the US, including Anthropic's own non-citizen employees.

Read that sentence carefully. The US government ordered an American AI company to block:

• Every European company using Claude Fable 5 or Mythos 5
• Every European researcher relying on these models
• Every European citizen who had subscribed
• Every non-American employee of Anthropic — including those working in San Francisco

Anthropic said it received the directive Friday evening. The Claude maker said it chose to shut down access entirely, given that selective compliance would have required blocking a wide swath of users — among them Anthropic's own foreign-born staff.

Anthropic faced an impossible choice: implement nationality-based discrimination against its own employees and customers, or shut down the models entirely. It chose the latter. Both options are catastrophic for European users.

The Letter — And What It Did Not Say

Commerce Secretary Howard Lutnick sent Anthropic CEO Dario Amodei a letter outlining the restrictions. Anthropic said the letter offered no explanation of the underlying national security rationale.

No explanation. The most powerful regulator in American technology policy ordered an American company to lock out every non-American user in the world — and provided no reasoning. This is the new normal of American AI export control: decree first, justify never.

Anthropic added that it believed the restriction stems from a narrow potential jailbreak — a capability available in other publicly deployed models. The asymmetry is revealing. OpenAI's GPT-5.5 and Google's Gemini 3.5 have comparable capabilities. Neither is subject to a foreign-national export ban.

Only Anthropic — the company that refused to waive safety guardrails for Pentagon warfare applications in February 2026 — has been singled out.
The pattern is consistent. Refuse the Pentagon, lose government contracts. Refuse the export directive, lose your business. The asymmetry between Anthropic and its US-government-aligned competitors is now policy.

What This Means for Europe — Six Concrete Consequences

One — Every European enterprise using Claude faces immediate continuity risk. Companies that built workflows on Claude Fable 5 or Mythos 5 have lost access overnight. Migration to alternatives requires re-engineering work that was not budgeted, scheduled or anticipated.

Two — European AI procurement assumptions are invalidated. Every European enterprise that selected Claude based on its safety credentials and capability profile now knows that American government policy can disable access without notice, without explanation, and without remedy.

Three — The EU AI Act enforcement timeline becomes structurally complicated. Brussels must enforce its transparency rules against AI providers whose access to European users is now controlled by a foreign government. The legal jurisdiction of the EU AI Act has not changed. The practical capacity to deploy regulated AI in Europe just did.

Four — European AI sovereignty is no longer a theoretical concern.
The European Commission's repeated calls for AI sovereignty — funded inadequately and pursued without urgency — are now vindicated by an event that the institutional response apparatus is not yet prepared for.

Five — Anthropic's own employees face an unprecedented workplace situation. Non-American employees at Anthropic — including those born in EU member states — cannot use the products their company builds. They are foreign nationals first, employees second, under the new American legal framework.

Six — Every other American AI provider is now subject to the same precedent. What happened to Anthropic on Friday can happen to OpenAI on Monday, to Google on Tuesday, to Microsoft on Wednesday. The legal infrastructure for selective American AI export control is now operational.

The Pattern — What We Have Been Tracking

gafam.ai has documented every step of the regulatory pattern that produced Friday's directive:
In February 2026, the Pentagon blacklisted Anthropic over safety guardrails. Earlier this month, the Pentagon began testing OpenAI and Google models to replace Claude in classified systems. Six days ago, we reported that Trump, Sanders and Altman all support government equity stakes in OpenAI. Eighteen days ago, we documented that Musk, Zuckerberg and Sacks killed Trump's AI safety executive order in three phone calls.

The dots connect into a coherent strategy: American AI policy is being shaped by an administration that rewards companies aligned with its preferences and punishes companies that resist. Anthropic resisted. Anthropic is being punished.

European users are collateral damage in an American political conflict in which Europe has no voice.

What Anthropic Itself Has Said About This

Anthropic has a coherent policy theory. In its Policy on the AI Exponential, Anthropic argues that governments should have legal authority to block or deter dangerous deployments, with transparency, independent evaluation, and safeguards against overreach.

Anthropic believes governments should have this authority — with transparency, independent evaluation and safeguards. The Lutnick letter provided none of those. The very company that argued for government authority over AI is now experiencing that authority applied without the safeguards it explicitly advocated for.

The lesson is sobering: even the AI company most committed to responsible government oversight cannot guarantee that government oversight will be responsible.

The European Perspective