One Week Later, Europe Still Cannot Use Anthropic’s Best AI. The Silence Speaks.
In a nutshell
Seven days ago, we reported that the US Commerce Department had locked European users out of Anthropic's most capable AI models. At the time, the open question was how long the shutdown would last.
We now have a partial answer: at least a week. And counting.
Where Things Stand Today
The models remain suspended. Anthropic's own status page still reads the same as it did a week ago: access to Claude Fable 5 and Claude Mythos 5 is suspended, with the company apologising for the disruption and working to restore access as soon as possible.
AWS confirmed the scope of the shutdown. To support compliance with the US Government export control directive, Anthropic asked AWS to revoke access to Claude Fable 5 and Claude Mythos 5 for all users — while all other models, including Opus 4.8, remain unaffected.
The resolution remains stuck in negotiation. Ongoing high-level talks with the White House reflect unresolved tensions over potential guardrail bypasses and model power, while the company publicly calls the move a likely misunderstanding and works toward restoration.
One week. No restoration. No public explanation. No timeline.
The Market Is Betting on the Outcome
When official channels go silent, markets fill the vacuum. A Polymarket prediction market has emerged specifically to trade on whether Anthropic will restore Fable 5 access — and the volume is significant. As of June 18, more than $770,000 had traded on the question of whether Claude Fable 5 would be restored to customers by specified dates.
That a prediction market exists at all is revealing. The restoration of a commercial AI product is now a speculative event — traded like an election outcome or a central bank decision.
The uncertainty that the Lutnick directive created has a dollar value, and the market is pricing it in real time. Traders are watching for any policy carve-out or technical fix that could quickly reopen access amid competitive pressure from other frontier AI labs.
The Detail Washington Has Not Explained
The technical justification for the ban remains contested. Anthropic's position is that the export control stems from a narrow jailbreak vulnerability — one that, by the company's account, exists in other publicly deployed frontier models that have not been subjected to equivalent restrictions.
If that account is accurate, the question becomes unavoidable: why Anthropic, and why now? The pattern we have documented over the past weeks offers a partial answer. Anthropic refused to waive safety guardrails for Pentagon warfare applications in February. The Pentagon began replacing Claude with OpenAI and Google models in classified systems this month.
Amazon — Anthropic's own investor and board member — triggered the Commerce Department directive through a phone call to the Treasury Secretary, as the Wall Street Journal, Reuters and The Information all reported.
One week later, none of these dots have been disconnected. They still form the same line.
What This Means for Europe — A Week of Lived Consequences
When we wrote about this a week ago, the consequences were prospective. Now they are lived reality.
Every European enterprise that built a workflow on Claude Fable 5 has spent a week without it. Every European researcher who relied on Mythos 5 has spent a week working around its absence. Every European subscriber who paid for access has spent a week paying for a product they cannot use. And Anthropic's own non-American employees have spent a week unable to use the models their company builds.
A week is no longer an emergency. A week is a new operating condition. The distinction matters: enterprises can absorb a one-day outage as an incident. A one-week outage with no restoration date becomes a procurement signal — a data point that reshapes how European organisations evaluate American AI dependency going forward.
The Quiet Workaround — Fable 5 Still Runs, For Some
There is a notable asymmetry in how the shutdown has played out. Even as access was revoked for compliance, Anthropic and AWS continued to describe Claude Fable 5 as available on Amazon Bedrock in US East and Europe Stockholm regions for approved configurations.
The picture on the ground is fragmented — full public access suspended, while specific enterprise and regional channels operate under different conditions. LLM Leaderboard
This fragmentation is itself a consequence of the directive. A model that was universally available on June 9 is now available through a patchwork of regional and contractual exceptions that most users cannot easily navigate. Clarity has been replaced by case-by-case access. For European enterprises, the practical effect is uncertainty — the most expensive condition in any procurement environment.
The European Perspective
One week into the Anthropic shutdown, the most important development is the absence of one: no restoration, no clear timeline, no public explanation from Washington. That silence is itself the lesson European policymakers should absorb. The EU AI Act regulates how AI may be deployed in Europe. It does not guarantee that American AI will remain available to deploy. Those are different forms of control, and the past week has demonstrated which one currently matters more for European operational continuity. A prediction market trading three quarters of a million dollars on whether a commercial AI product will be switched back on is not a functioning market for enterprise infrastructure. It is a symptom of a dependency that has become a vulnerability. The European response cannot be another diplomatic protest. It must be the unglamorous, expensive, multi-year work of building AI infrastructure that no foreign directive can switch off — Mistral, Aleph Alpha, Black Forest Labs, the European AI Factories programme. One week ago that argument was theoretical. Today it has a price tag, a prediction market, and seven days of European downtime attached to it. gafam.ai will be watching.
We are not first. We are right.