Europe's AI Sovereignty Debate Returns After Anthropic Outage
European founders and investors spent TechBBQ 2026 debating who actually controls AI after June's Anthropic outage showed how quickly access can disappear. Signal's president called agentic AI a data collection apparatus, and speakers urged local, open, and sovereign options. Here is what was said and what it means for private AI users.
On this page
Copenhagen kept asking who controls AI
TechBBQ, the annual Nordic tech conference in Copenhagen, wrapped up on 27 August with a theme about agency, but the question delegates kept circling was narrower: who is actually in control? TechCrunch's reporter on the ground found that panel after panel, and plenty of hallway conversation, kept returning to AI sovereignty, meaning not which model writes the best prose, but who holds power when a provider changes its mind [1].
The mood was practical rather than ideological. One startup executive said the Anthropic disruption had seriously disturbed his software team's work. Another participant argued that dependence on US and Chinese AI providers could someday cause Europe severe problems, while conceding that "for now, everything is still pretty much OK" [1].
The outage that turned sovereignty into a practical question
The debate has a concrete trigger. On 12 June, Anthropic disabled access to its Claude Fable 5 and Mythos 5 models to comply with a US government export-control directive [2]. Wired reported that the models had launched only days before they were pulled [4]. European politicians reacted to the halt by calling it a wake-up call and warning about Europe's deep dependence on a handful of US-based AI firms [3].
That detail reframes the sovereignty discussion for ordinary users. The cutoff was not a price rise or a deprecated feature. Access was removed at government speed, for brand new models, and no customer contract protected them. Founders who build on a single provider now treat that episode as a live case study in concentration risk [1].
Signal's president: agents are a data collection apparatus
Meredith Whittaker, Signal's president, joined a panel on privacy and AI and criticized the push to embed AI agents directly into operating systems, citing ChatGPT's iMessage integration as an example. She described the current era as a data collection apparatus, argued that AI lab marketing distracts from the collateral consequences of mass data collection, and told the room: "There is still a huge market need for privacy" [1].
Emad Mostaque, co-founder of Stability AI and Intelligent Internet, offered a harder-edged definition of the sovereignty idea. He called sovereignty "the ability to resist power being exerted over you", predicted that power will keep concentrating in a few AI labs, and summed up his view bluntly: "the person that controls the AI controls the country" [1].
What this means for local and private AI
For readers deciding where their AI should run, none of this is abstract. A cloud subscription can disappear because of a government directive, an acquisition, or a contract clause, and none of those events ask your permission first. Local and self-hosted models fail differently: they keep working when a provider pulls access, and they keep your prompts off someone else's servers.
That is why the European conversation sounds so much like the on-device pitch. Control, continuity, and data that stays put are exactly what local AI offers, and exactly what the June outage put in doubt [1]. The practical takeaway for individuals mirrors the one emerging in Copenhagen: keep a local fallback model on your devices, check where your provider processes prompts, and treat dependence on a single AI vendor as a risk to manage rather than a convenience to ignore.