Europe’s Coordination Moment

Governance in the age of AI infrastructure

A policy bridge from the series “Examining Altman’s Moore’s Law for Everything” — on the political economy of cognitive abundance.

“The interface was never the first instinct. It was the hard-won conclusion.”

Every time a new coordination system has emerged that institutions cannot fully own, the same arc has tended to unfold: attempted control, institutional strain, then a negotiated interface. The Hanseatic League. The Lex Mercatoria. The joint-stock company. Each produced an interface between legacy governance and emergent coordination. The interface was never the first instinct. It was the hard-won conclusion — the moment when the costs of resisting coordination exceeded the costs of governing it.

That pattern is repeating. The question is whether the interface gets built deliberately or forced by crisis.

The urgency of this moment is not lost on those building the technology. At the World Economic Forum in January 2026, the CEOs of Google DeepMind and Anthropic both called for new governing institutions — an IAEA-equivalent for AI, international coordination mechanisms, safety-first oversight. “There’s a huge amount of responsibility — probably too much — on the people leading this technology,” said Demis Hassabis. Dario Amodei warned of high GDP growth alongside potentially high unemployment, and acknowledged uncertainty about what AI sovereignty even means. The series this piece draws on accepts the urgency of that call. What remains unspecified — across both the manifesto that prompted the series and the field’s own leaders — is the institutional form that addresses the participation and legitimacy layer underneath the distributional and safety questions.

The deeper structural problem this series has traced is not regulatory lag. It is a gap that opened when the cost of horizontal communication collapsed, allowing activity and consequences to travel across boundaries that responsibility did not. The internet made communication cheap. It did not solve what coordination actually requires: verified identity, consequence for defection, aligned incentives. It also reduced many of the institutional costs associated with defection, and offered few mechanisms to make defection costly at scale.

Each wave of technological abundance appears to reveal the next layer of scarcity: information abundance shifted value toward attention; communication abundance shifted it toward trust; intelligence abundance may shift it toward coordination itself — the capacity to act on shared understanding without defaulting to coercion or paralysis. AI intensifies the asymmetry: the same infrastructure that makes content cheap makes verification expensive, and the infrastructure required to perform trust-grade verification is increasingly concentrated in the same systems whose optimization functions we are already questioning.

What this moment may require is not primarily a new regulation. It may be a governance interface — the institutional membrane between democratic authority and coordination systems democratic institutions did not build and cannot fully own.

A June 2026 study by Project Liberty Institute and the Global Solutions Initiative, drawing on consultations with 13 middle power governments across five continents, found that the primary constraint on building digital infrastructure adequate to the AI era is coordination, not ambition — and that no middle power can build sovereign AI capacity alone.

That interface has three structural requirements, each addressing a different layer of the coordination failure.

The first is an institutional intermediary capable of performing a double function: acting as an airlock between participants and large-scale digital infrastructures, and making participation institutionally legible. Hernando de Soto described dead capital as assets that exist but cannot be mobilized because institutions cannot verify them. In a digital economy, relational reliability, contribution history, and coordination capacity risk becoming the new dead capital — real, consequential, and invisible to the systems allocating resources. The intermediary renders operational relationships institutionally legible without surrendering them to extractive capture. Its authority is architectural, not moral.

The same structural problem runs through the redistribution debate. Labor historically performed four functions simultaneously — distributing purchasing power, providing bargaining leverage, conferring dignity, and securing civic participation. A universal dividend may preserve the first while leaving the others structurally unaddressed. Income is not the same as standing. The governance architecture the shared floor requires must address both.

The second requirement follows directly: a relational layer that treats verified human relationships as the unit of social capital. A bot farm can generate ten thousand synthetic identities. It cannot sustain ten thousand reciprocal, historically validated relationships in which real people put social capital at risk for one another. The friction of maintaining accountable relationships is not inefficiency — it may be the most important governance signal available. The relational layer enables mutual accountability without collapsing into centralized social scoring: pluralism preserved, reciprocity visible, defection costly.

The third is not a choice between centralization and decentralization, but the institutional capacity to navigate between them. The 1953 North Sea flood revealed what the decentralized Dutch water boards had not adequately anticipated: local legitimacy is essential and insufficient when the threat exceeds local scope. The political response — the Delta Works and a restructured governance architecture — federated local boards upward without abolishing them. Local accountability remained local. Crisis synchronization became national. Digital governance may require the same design: subsidiarity in ordinary operation, defined escalation under stress. These are not contradictory properties. They require explicit architecture.

These three structural requirements translate into three operational conditions for a shared floor — and one precondition that must come before all of them.

The precondition is a governance structure that makes reliance on the interface warranted. Before interoperability standards or procurement frameworks, actors need enforceable assurance that the shared layer cannot be captured by any single state, bloc or firm. This requires legal obligations to the commons itself, together with transparent governance, contestability and protection against capture. Without that architecture, the floor is not shared. It is rented.

• The first operational condition is technical interoperability — open standards, API commitments, data portability mandates. These are the roads and ports of the historical interface: infrastructure the sovereign layer provides so that coordination networks can function across it rather than around it. They are necessary. But open standards without governance is a commons without management, and data portability without fiduciary protection risks portability into capture — moving data freely between systems that all optimize against the person who generated it.
• The second is governance that scales under stress. Subsidiarity for ordinary operation; federated escalation under crisis. Decision-making distributed for legitimacy; coordinated for resilience. These are not in tension — but they require explicit design in advance, not improvisation after the dike breaks.
• The third is economic sustainability. The shared floor is not a market product. It is a commons with a business model. Procurement is the lever that simultaneously shapes markets, signals legitimacy, and scales infrastructure before it is complete. Member states spend over two trillion euros annually. Coordinated procurement commitments could create demand before the technology is ready. Contribution models calibrated to GDP or population — rather than raw spend — could protect smaller economies from being priced out of governance itself.

Europe’s position here is structural, not merely political. The European Union is itself an imperfect but real institutional interface between 27 sovereign states — multi-jurisdictional governance without a single sovereign, coordination among divergent interests, decision-making that must balance local legitimacy with collective resilience. That institutional achievement is undervalued as a governance template.

Get Gael Van Weyenbergh’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

Europe has spent a decade regulating digital infrastructures it did not build and does not meaningfully control. Regulation alone cannot create infrastructural leverage, nor does the existence of individual remedies ensure that recurring consequences alter institutional action. The problem is not the absence of rules. It is the missing interface between regulatory authority, operational systems and the people living with their consequences.

The problem is infrastructural dependency — and the leverage to address it is physical, not only legal. Data centers need land, security guarantees, and energy. Those are European assets, and their terms can be conditioned. Offering hyperscalers favorable hosting conditions in exchange for contractual access guarantees to frontier capabilities is not accommodation. It is statecraft.

If 27 states can converge on even a partial shared floor, the demonstration effect may be global. Like-minded middle powers in Africa, Latin America, and Southeast Asia face the same sovereignty dependency problem and lack the leverage to solve it individually. Europe converging first could create a template others adopt, adapt, or federate with. The coalition required is not ideological alignment. It is interface compatibility.

Europe may not need to own the global commons. It may need to build the interface that makes federating with it the rational choice.

Four questions this piece does not answer — but which require an institutional home.

First: what is the legally binding circuit breaker that shifts authority from distributed legitimacy to crisis coordination when digital infrastructure fails at millisecond speed — and who controls it without that control becoming permanent?

Second: which existing procurement instrument can be repurposed as the shared floor’s financial vehicle, and what minimum legal framework prevents any single member state from capturing what was built collectively?

Third: how does civic legibility survive when a state classifies the shared infrastructure as a strategic asset — when the coordination layer is claimed by intelligence services and citizens risk being reduced to data nodes or administrative units in a classified system?

Fourth: several sovereign primitives are already migrating from institutional accountability to private sector logic — not through legislation but through the ordinary accumulation of technical capability.

• Identity — three billion people authenticate daily through Apple ID or Google login, private systems functioning as de facto civil registries.
• Currency — Meta’s planned stablecoin integration would compress payment infrastructure, identity, and behavioural data into a single commercial platform without fiduciary accountability to users.
• Cognitive capacity — the Government of Malta has contracted OpenAI to provide ChatGPT Plus to all its citizens, with a former G7 finance minister negotiating AI access with European governments as enterprise clients.
• Legal arbitration — when Apple, Google, and AWS simultaneously deplatformed Parler within 24 hours, three private firms exercised the functional equivalent of economic exile without judicial process or right of appeal.
• Compute infrastructure — three hyperscalers (AWS, Azure, Google Cloud) now hold more processing capacity than most national governments, the physical substrate on which every other sovereign function runs.
• Defense and intelligence — Alex Karp’s “Digital Republic” essay argues explicitly that private firms should hold sovereign intelligence functions; Palantir’s battlefield awareness infrastructure already rivals state intelligence agencies in operational reach.
• Security and frontier AI capability — Anthropic’s Mythos deployment, restricted to US territory and a vetted circle of national security actors, demonstrates that the most consequential cognitive infrastructure is already becoming geopolitically gated.

When the private sector performs institutional functions without inheriting institutional accountability obligations, the democratic mediation gap is not a future risk. It is the present condition. Which of these primitives is most tractable for European intervention — and which has the highest leverage for demonstrating that the interface can be governed rather than merely regulated?

These are not objections to the architecture. They are its next layer. Europe has the institutional capacity to begin answering them.

The interface will emerge either deliberately or under stress. The remaining question is under which conditions.

→ The four-part series this piece draws on: Examining Altman’s “Moore’s Law for Everything”
→ Europe in the Coordination Age series
→ The historical interface pattern: The Handshake That Didn’t Happen Yet · When Institutions Meet a Newcomer, the Interface Is What Survives
→ Explore the full governance architecture: Trust as Infrastructure

Written with AI assistance. Ideas and arguments by the author.