EU Officials Flew to D.C. Seeking Access to Anthropic Models

The European Union flew a senior commissioner to Washington this week to ask the Trump administration — not Anthropic — to turn its frontier AI models back on. Commerce's June 12 export order cut Europe off from Claude Fable 5 and Mythos 5. Thirteen days later, the models are still dark, OpenAI's rival cyber model is live, and Brussels has confirmed the kill switch sits in Washington.

What happened: European Commission Executive Vice President Henna Virkkunen discussed restoring access to Anthropic's most advanced models during a Washington trip this week, Bloomberg reported June 25. The bloc's route to the technology runs through the White House — not the San Francisco lab that built it.

The cutoff:

• On June 12, Commerce Secretary Howard Lutnick ordered Anthropic to block Fable 5 and Mythos 5 for any foreign national, inside or outside the United States.
• Anthropic could not verify user nationality at inference time, so it disabled both models for everyone — including U.S. citizens.
• Less capable Claude models, including Opus 4.8, remain online.
• The order arrived three days after Fable 5's public launch and reportedly followed a jailbreak concern flagged by Amazon researchers.

Why Europe cared: Weeks before the ban, ENISA — the EU's cybersecurity agency — had become the first European organisation admitted to Project Glasswing, Anthropic's cyber-defender coalition. NATO joined in the same expansion. European governments and companies that gained Mythos access lost it with no notice and no return date.

What Brussels said:

• Virkkunen, on social media June 15: "Europe represents an economic opportunity, not a security risk. We are and will remain a trusted partner."
• Commission spokesperson Thomas Regnier: contingency measures "should not be discriminatory against partners" and the episode "underlines the need for Europe's technological sovereignty."
• MEP Christophe Grudler (Renew/FRA): "The United States is once again demonstrating... that the US holds a real 'kill-switch' over essential technologies and that they are more than willing to use it."
• MEP Bart Groothuis (Renew/NLD): "Europe needs its own LLMs and open weight models or face digital colonisation."

The UK got the same answer: Downing Street pressed the White House for a carve-out and was told there was "zero chance," according to The Next Web — a refusal that made clear allied capitals cannot negotiate terms for a private American product.

The thaw that isn't: Anthropic CEO Dario Amodei met President Trump at the G7 summit in France. Trump later told Axios he no longer views Anthropic as a national security threat. Fable 5 and Mythos 5 remain offline. Warm words are not a product release.

The fairness gap: On June 22, OpenAI launched GPT-5.5-Cyber — scoring 85.6% on the CyberGym benchmark versus Mythos 5's 83.8%, per OpenAI and Decrypt. OpenAI ran pre-deployment tests with federal agencies including CAISI and the Office of the National Cyber Director. Nobody ordered it offline. Seoul Economic Daily noted the regulatory asymmetry: the lab Washington dimmed is not the one shipping defenders their ammunition.

Convina's view: This is what AI geopolitics looks like when export law meets cloud APIs — allies discover they are foreign nationals in someone else's security regime. Brussels is right that Europe is not China's military-industrial complex. Washington treated the distinction as irrelevant anyway. Every enterprise, bank, and cyber agency betting on American frontier models should update its risk register: vendor continuity is now a diplomatic variable, and your competitor may hold the same capabilities under a friendlier political profile. Europe will fund Mistral harder after this. It still cannot replicate Mythos by press release.