The question hanging over Copenhagen during this year’s TechBBQ conference was not about what artificial intelligence can build, but who gets to decide what it builds and for whom. Across the Nordic tech summit, which wrapped on August 27, the conversations among investors, founders, and operators kept circling back to a single, urgent concern: how Europe can reclaim control over the technology powering the AI wave, instead of renting that power from the United States and China.
That debate over sovereignty — which aligned neatly with the conference theme of “Emerging from Agency” — was given fresh urgency by an incident earlier this year when Anthropic made its AI models Mythos and Fable unavailable to users outside Europe. For some in the European ecosystem, the move was a wake-up call. One startup executive told me the sudden loss of access seriously disrupted his software team. Another shrugged it off, acknowledging that relying on two other geopolitical powers for AI might one day cause Europe severe headaches, but for now, everything remains functional enough.
The divergence in reactions captures the tension running through the European tech scene. Europe knows it needs its own AI infrastructure. It knows the models and the compute power underpinning this wave are largely controlled elsewhere. But acting on that knowledge — building the alternatives, funding the bets, accepting the trade-offs — is a different matter entirely.
The Mythos and Fable Incident: A Sovereignty Stress Test
When Anthropic pulled its Mythos and Fable models from non-European markets earlier in 2026, the move struck many in the European AI community as a preview of things to come. If a major AI lab can withdraw access to its most capable models on short notice, what happens when that access is tied to geopolitical shifts, regulatory disputes, or commercial disagreements?
For one European startup executive, the answer came quickly. His team had been relying on these models for production workloads. When access was cut, workflows broke, timelines slipped, and the company had to scramble for alternatives. The disruption was real, and it left him questioning whether any European company building on top of American or Chinese foundation models could ever truly de-risk its business.
Another executive took a more measured view. Yes, the incident was concerning in theory, but in practice, Europe still has access to a wide range of capable models. The immediate pain was limited. The real danger, he argued, is not today’s inconvenience but tomorrow’s dependency — the slow accretion of reliance on infrastructure that Europe does not own and cannot control.
That tension between present pragmatism and future risk defined much of the debate at TechBBQ. No one disputes that Europe needs more sovereignty over AI. The question is whether the will to build it matches the urgency of the moment.
What AI Sovereignty Actually Means for Europe
To understand the sovereignty debate at TechBBQ, it helps to distinguish between different layers of control. There is ownership of the models themselves — the weights, the architectures, the training data. There is ownership of the compute infrastructure — the chips, the data centers, the energy supply. And there is ownership of the data that feeds the whole system. Europe has strengths in some areas and glaring gaps in others.
Ellen de Brever, an angel investor and head of partnerships at the Novo Nordisk Foundation Cellerator, captured the mood when she said the conversation at TechBBQ had shifted from what AI can do to what people are actually willing to let it do. “TechBBQ’s theme of agency felt particularly timely this year,” she told me. “The debates weren’t about smarter machines, but about human judgment and who gets to stay in the driver’s seat.”
De Brever pushed the point further. “The age of AI agents isn’t really a story about machines gaining agency,” she said. “It’s a story about humans deciding what to give up.” That framing reframes the sovereignty question: it is not just about building European AI, but about deciding which aspects of human decision-making should be ceded to machines at all, regardless of where those machines are built.
Meredith Whittaker’s Warning on Privacy and AI Agents
One of the most pointed sessions at TechBBQ was a panel featuring Signal President Meredith Whittaker, which I moderated. The question on the table was whether privacy and AI can coexist. Whittaker’s answer was blunt: not without serious structural changes to how AI is being integrated into everyday technology.
Whittaker singled out the integration of AI assistants and agents into operating systems — pointing to what ChatGPT is doing with iMessage as a prime example. She argued that this AI era is creating what she called a “data collection apparatus,” and that AI labs are using clever marketing to make people forget about “the collateral consequences” of acquiring so much data.
“There is still a huge market need for privacy,” Whittaker said. “And particularly with the sovereignty concerns, we will have customers here.”
Her point was that sovereignty and privacy are not separate issues. If European users are feeding their personal and professional data into AI systems controlled by American or Chinese companies, then Europe is exporting not just its data but its capacity to make independent decisions about how that data is used. Sovereignty, in this view, starts with data governance.
Emad Mostaque on Power, Control, and the Agentic Workforce
Emad Mostaque, co-founder of Stability AI and Intelligent Internet, joined me on a panel about how an agentic workforce will impact the economy and what that means for personal sovereignty. His framing was characteristically direct.
“Sovereignty is the ability to resist power being exerted over you,” Mostaque said. He spoke about the concentration of power in the hands of a few AI labs and drew a stark conclusion: “Inevitably, every country will be run by AI, and the person that controls the AI controls the country.”
That line resonated through the conference. Mostaque was not making a prediction about a distant future. He was describing a dynamic already in motion. If the most capable AI systems are built and controlled by a handful of companies in two countries, then every other nation is operating on borrowed intelligence. The implications extend far beyond technology into geopolitics, economic competitiveness, and democratic governance.
Mostaque’s panel explored what happens when AI agents begin performing cognitive work at scale, complementing the physical work increasingly done by robots. The combination points toward something much larger than another technological revolution.
Rethinking Work, Democracy, and Participation
Mia Negru, advocacy and engagement director for the nonprofit Life With Artificials, sent me a message after the Mostaque panel that crystallized the stakes. “If intelligent agents perform cognitive work and robots increasingly perform physical work, we may be approaching something much bigger than another technological revolution,” she wrote. “We may have to rethink ownership, work, democracy, economic participation and eventually even who gets to participate in society.”
Negru’s message captured why the sovereignty debate at TechBBQ felt different from previous years. It was not just about technology strategy or venture capital allocation. It was about the kind of society Europe wants to be — and whether it will have any say in the matter.
Europe’s AI Infrastructure Gap: The Compute Problem
Underlying the sovereignty debate is a hard infrastructure problem. Training and running frontier AI models requires massive compute resources — specialized chips, data centers, energy — and Europe currently lacks domestic capacity at scale. The continent has world-class research institutions and talented engineers, but when it comes to the raw computational power needed to compete with American and Chinese labs, Europe is playing catch-up.
Several attendees told me that the conversation about compute access was a recurring theme in private discussions. European founders building AI-native products are acutely aware that their ability to scale depends on cloud infrastructure provided by American companies. Some are exploring partnerships with European data center operators. Others are watching European Commission initiatives on sovereign cloud and high-performance computing with cautious optimism.
But the gap remains large, and it is growing. The latest generation of AI models requires clusters of tens of thousands of specialized accelerators. Building and operating those clusters is extraordinarily capital-intensive. No European company has yet committed the kind of resources that OpenAI, Google, Meta, or Anthropic have poured into their infrastructure.
The question of whether Europe needs to build its own frontier models or can achieve sovereignty through a combination of open-source models, specialized applications, and strategic compute investments was hotly debated — but not resolved.
The Conference Beyond the Stage: Connections and Caution
When the panels were not in session, the conversation continued at happy hours and rooftop gatherings across Copenhagen. Lovable, Nvidia, OpenAI, AWS Startups, and HSBC co-hosted a rooftop barbecue where I spoke with founders about Denmark’s growing tech hubs. London’s Ada Ventures hosted a gathering at the cocktail bar Bird, where I chatted with a health-tech founder about increased innovation in the women’s brain health space.
The speakers’ dinner took place on an island in the middle of Copenhagen, where the international press sat together and watched fire dancers. The after-party continued upstairs, with everyone tempted to stay for one more drink until the bar tab closed.
These moments might seem peripheral to the weighty questions of sovereignty and control. But they are where relationships form, where trust builds, and where the informal networks that sustain a tech ecosystem are strengthened. De Brever put it well: “In a room full of conversations about machines, the most memorable moments still come from the simplest thing. Human connection, face-to-face, building relationships, sharing ideas, and reminding each other of what technology can never quite replace.”
How TechBBQ’s Theme of Agency Reframed the AI Debate
The conference theme of “Emerging from Agency” was not accidental. It was chosen to provoke exactly the kind of conversation that unfolded. Agency — the capacity to act independently, to make choices, to exert control — is precisely what is at stake in the AI sovereignty debate.
For Europe, the path to agency runs through several interconnected domains. There is the regulatory domain, where the European Union’s AI Act creates a framework that could either foster innovation or constrain it. There is the investment domain, where venture capital and corporate R&D budgets must flow toward domestic AI infrastructure. There is the talent domain, where European researchers and engineers need reasons to build their careers at home rather than migrating to Silicon Valley or Shenzhen.
And there is the public domain, where citizens need to understand what is at stake and what is being asked of them. Sovereignty is not just a project for policymakers and founders. It requires societal buy-in — an acceptance that building European AI capacity will require trade-offs, investments, and perhaps a slower pace than importing ready-made solutions from abroad.
The debate at TechBBQ made clear that the European tech community is not unified on how to proceed. There are optimists who believe Europe’s strengths in vertical AI applications, life sciences, and industrial automation will give it a differentiated and defensible position. There are pessimists who argue that without frontier models and massive compute, Europe will be permanently subordinate. And there are pragmatists who say the answer lies somewhere in between — building selectively, investing strategically, and accepting that full sovereignty may be an aspiration rather than a destination.
What the Rest of the World Can Learn from Copenhagen
The conversations at TechBBQ have implications beyond Europe. The sovereignty question applies to every country that is not the United States or China. Japan, India, Brazil, the nations of Southeast Asia and Africa — all face the same fundamental choice: build your own AI capacity or accept dependency on a small number of external providers.
Several panels at TechBBQ directly addressed this global dimension. There was a summit on how the Nordic and African ecosystems can collaborate on innovation and investment, with participants exploring whether Europe’s regulatory approach and Africa’s mobile-first infrastructure could complement each other. There were discussions about how to get more growth capital into Europe, with comparisons to the funding ecosystems in other regions.
What emerged from these conversations was a recognition that sovereignty is not inherently zero-sum. One region’s investment in AI capacity does not necessarily come at the expense of another’s. In fact, a more distributed AI ecosystem — with multiple centers of gravity — might be more resilient, more innovative, and more aligned with democratic values than one dominated by two powers.
The Practical Consequences of Renting Intelligence
For European companies building on American or Chinese foundation models, the risk is not hypothetical. Every API call, every fine-tuning run, every deployment creates dependency. If the provider changes its terms of service, its pricing, its geographic availability, or its model behavior, the downstream customer bears the cost.
This is not a new problem. The software industry has dealt with vendor lock-in for decades. But AI introduces a new dimension: the models themselves are opaque and constantly evolving. A company that builds a product on top of a particular model cannot fully predict how that model will change, how its performance will shift, or what limitations will emerge.
Several founders at TechBBQ told me they are already taking steps to reduce this risk. Some are fine-tuning open-source models as alternatives to proprietary ones. Others are building abstraction layers that allow them to swap out underlying models without rewriting their applications. A few are investing in their own small-scale training infrastructure, targeting specialized domains where they believe European expertise can prevail.
These are sensible precautions, but they are not a substitute for structural sovereignty. They reduce the pain of dependency without eliminating the dependency itself.
The most forward-looking conversations at TechBBQ were not about today’s workarounds but about tomorrow’s infrastructure. What would it take for Europe to have its own frontier model, trained on European compute, governed by European norms? No one had a complete answer, but the question itself felt newly urgent — and newly possible.
That shift in sentiment may be the most important outcome of the conference. A year ago, the idea of European AI sovereignty was a fringe concern, raised mainly by policy experts and a handful of vocal founders. At TechBBQ 2026, it was the central conversation. The terms of the debate have changed, and while the path forward remains uncertain, the direction is clear: Europe is no longer asking whether it needs AI sovereignty, but how to achieve it — and what it is willing to give up along the way.
