Skip to content

AI

Europe's AI crowd went to a BBQ and left asking who holds the leash

At TechBBQ, the hallway chatter kept circling one question: who actually controls the models? Anthropic's Mythos and Fable going dark outside Europe turned a theme slide into a real worry.

TL;DR

  1. TechBBQ's theme, 'Emerging from Agency', landed uncomfortably well on AI sovereignty.
  2. Anthropic's Mythos and Fable became unavailable to users outside Europe earlier in 2026.
  3. Startups worry that renting AI from the US and China brings geopolitical risk.

What happened

The sovereignty question hit harder because of recent history. Earlier this year, Anthropic's models Mythos and Fable became unavailable to users outside Europe. For a lot of people in the European scene, that was the moment the abstract debate turned practical.

The incident pushed many to think seriously about owning the models and the infrastructure under this AI wave, instead of renting that power from the US and China.

Not everyone panicked. One startup exec said the episode seriously disrupted his software team. Another shrugged it off.

The shrugger did concede the long-term point. Leaning on two other geopolitical powers for AI might one day give Europe severe headaches. For now, he said, things are still pretty much OK.

Ellen de Brever, who invests as an angel and runs partnerships at the Novo Nordisk Foundation's Cellerator, said the conversation has shifted. It is moving from what AI can do to what we are willing to let it do.

Why it matters

Renting intelligence by the API call is cheap and fast. It is also a dependency, and dependencies have owners.

The debate moved from what AI can do to what we are willing to let it do.

The conference mood suggests that dependency is no longer just a policy-paper topic. When a model you build on can vanish from your region, "vendor risk" stops being a line in a spreadsheet. Even the exec who shrugged admitted the exposure is real, just not yet painful.

This is a single report from one conference, so treat it as a mood reading, not a market shift. But the mood is specific: European startups are worried that reliance on US and Chinese AI infrastructure could create future geopolitical risks.

What it means for Europe

Europe is both the audience and the subject. The ecosystem is weighing what it means to own models and infrastructure rather than borrow them, and the conference gave that argument a stage.

The practical split is already visible. Some teams were hit hard by the Mythos and Fable incident. Others barely noticed. That gap tells you who has built a single point of failure and who has not.

Sovereignty talk is cheap at a conference with a theme slide. The test is whether companies change how they buy AI before the next disappearing act.

What to do with it

  • Map your dependencies.List every model and AI service your product touches, and note where each one is hosted and who controls it.
  • Build an exit ramp.Make sure at least one critical workflow can switch to a second provider without a rewrite.
  • Ask vendors the awkward question.Get regional availability and continuity terms in writing before you sign.
  • Watch the sovereignty debate.If you sell to European enterprises, expect "who controls the model?" to show up in procurement questions.

Ciao, sapiens. Rent wisely: the landlord can change the locks.