Five big ideas for the 2026 State of the Union

Next week President von der Leyen delivers the State of the Union. The speech, and the Letter of Intent that goes with it, begin the process that becomes next year’s Commission Work Programme – which makes this the week to say what we think belongs in it. 

The context is critical. Europe has had an anxiety-inducing summer. At least 35,000 more people died across four back-to-back heatwaves than would normally be expected, and that is on incomplete figures covering barely half the continent. At the end of July tens of thousands crossed into Ceuta over two days, and at least 72 died doing it. The geopolitics of the Middle East and the supply constraints in the Straight of Hormuz have sent oil above $100 a barrel again with further pressure on European wallets and reminding us of the perils of our dependence on the fossil – rather than bio – economy. Challenging contexts always  create fuel for populists, and on Sunday the AfD took nearly 44% of the vote in Saxony-Anhalt – more than double the Chancellor’s party.

None of that is a technology agenda, but it is the ground any EU technology agenda has to stand on.

The recommendations below cluster around five big ideas: testing the big decisions against who inherits them; backing biotech as a strategic sector rather than a branch of health policy; extending cognitive integrity safeguards to all consumer AI, not only social media – and not only children; expanding compute fast, without handing the bill to the grid or the towns hosting it; and building the global AI rules alongside all the countries the superpowers treat as bystanders.

#1 Put future generations at the heart of the strategy

The Commission adopted its Intergenerational Fairness Strategy in March and it got less attention than it warranted. No other legislature is doing long-horizon work at this scale, and the EU should be less shy about saying so. When it landed we welcomed it with one reservation: Europe is finally thinking about the future, but not yet about how it intends to govern the technologies building it.

• Use the Strategy as the test for strategic decisions. In practice that means intergenerational impact assessments on the major technology, energy and economic files, weighing the risks of acting against the opportunity costs of not acting — the side of the ledger usually left blank. Our submission to the Strategy consultation sets out how.
• Implement it in full, including the Futures Balance Tool. Foresight earns its keep when it changes decisions rather than when it produces reports, which means futures literacy built into how officials work rather than a unit at the edge of the building.

#2 Go big on Biotech

Europe is understandably focussed on the digital tech stack, but in biotech, Europe has a distinct comparative advantage with an outstanding  industrial base, the labs and the talent. Yet biotech is still treated as a branch of health policy rather than a frontier technology sector in its own right.  This is a mistake.

Our Biotech team spent this year working through ten years of Horizon grants to European biotech firms. Strategic sector, generic funding, published a fortnight ago, finds that grants reach firms scaling established processes more readily than firms doing novel work, which is roughly the opposite of what a strategic sector policy is for.  The solutions are clear:

• Elevate biotech to a strategic priority and take the ambition past the Biotech Act. The aim is to hold the value chain democratically and with European sovereignty, which is what makes biotech a pillar of strategic resilience in a post-fossil economy rather than simply another growth sector. Our submission to the Biotech Act II call for evidence in June sets out what that involves. A European biotech stack is achievable inside this mandate.
• Fund shared infrastructure and move away from single-firm subsidies. The Chips Act is the precedent worth copying. Both the EU and US versions moved from subsidising single companies towards shared scale-up capacity bridging lab and production, and the audits since suggest it worked. For biotech that means three layers — physical, digital, and the knowledge infrastructure connecting them — funded publicly so the returns stay with European citizens.
• Deliver EU-INC. A single European start-up regime goes at market fragmentation, which is one of the genuine constraints on scaling here, and it is what allows our innovators to grow as Europeans rather than somewhere else. The Commission has already backed it; the European start-up community is calling for EU leaders to get EU-INC right in the next 100 days

#3 – Protect our cognitive integrity from the risks of AI systems

AI holds real promise for European science and industry, and we must capture the opportunities. But the consumer-facing systems citizens interact with every day are a different proposition: they shape attention, judgment and emotional response, not by chance – but by design.

We call this “cognitive integrity”, meaning the capacity of individuals and societies to think and make decisions autonomously. Freedom of thought is already specified in  Article 10 of the EU’s Fundamental Rights Charter. What is new is the scale at which it can be worn down, and what that accumulates into — a risk to European citizens’ ability to think for themselves, which in turn weakens democratic resilience and reduced competitiveness – as citizens are exposed for foreign interference and misinformation and consumers are nudged into accepting poor economic deals.

The cognitive integrity workstream launched this month; in our first report, we examine the psychological needs that pull people towards these systems, trace how business incentives, data and design choices shape their effects on people and society, and identify where existing EU governance falls short.

Europe is right to move quickly on children and teenagers. But the design practices in question exploit vulnerabilities we all share – in how our attention is captured, how our trust is won, and how our judgment can be displaced. People age out of protection, they do not age out of vulnerability. The Commission could do three things in response:

• Set the scope at AI systems, not social media. Europe’s approach so far targets social media platforms, and protects under-16s. Our research says that is too narrow on both counts. The scope should cover every consumer-facing AI system capable of shaping or weakening cognitive integrity – and the safeguards should apply wherever vulnerability runs high, which means age, but also circumstance, isolation, health and how much digital literacy someone happens to have.
• Treat cognitive integrity as an AI safety question and build the evidence base independently. Independent researchers, not just companies, should assess how people’s interactions with AI affect attention, behaviour, and decision-making across systems’ features and over time. That means moving beyond the current assessment paradigm that focuses on a system’s technical capabilities in isolation.
• Keep the legal framework fit for purpose. Assess regularly whether the instruments covering consumer-facing AI — the AI Act, the DSA, the DMA, consumer law — function coherently as a set, and whether they remain robust and adaptable enough to protect and empower citizens as the technology changes. Coherence is the weak point: the enforcement gap we have been writing about since 2024 is as much about instruments failing to work together as about any single rule going unenforced. Collective redress belongs in that review, so that no citizen has to prove alone what an algorithm did to millions.

#4 – Deliver the AI build-out within planetary and social limits

Europe is about to expand its computing infrastructure faster than at almost any point in its recent history. We will not win this by building the biggest data centres. The binding constraint is power, grid connection queues in the major hubs already run into years, and the advantage genuinely available to us lies in how we govern the build-out, because that determines how fast it can actually happen.

Three tests follow. Expansion should be fast enough to keep Europe competitive, fair enough to hold public consent, and conditional enough to stay consistent with commitments already made. The alternative is on display in the United States, where local opposition is now delaying projects that were waved through, and communities are demanding safeguards that would have cost far less to design in at the start. This is the question our Tech & the Planet work exists to answer. Three clear next steps are to:

• Apply the fast, fair and conditional test to the compute build-out, beginning with a single European methodology for assessing data centres under the Cloud and AI Development Act, covering energy, water, land and community impact, so assessments are comparable across members and siting decisions can be defended when challenged.
• Bring AI into the European Fair Transition Observatory, so we know early and with evidence where compute demand complements Europe’s energy and industrial transition and where it has started to compete with it. Trading one dependency for another is not sovereignty.
• Ask the European Scientific Advisory Board on Climate Change what an AI build-out consistent with Europe’s climate pathway actually requires, including under overshoot conditions. Better to have that number now than to discover it in 2032.

#5 – Be a global policymaker – not a policytaker

Beijing has called for extensive international cooperation on AI governance. American tech firms are pushing for a governance body of its own design, on its own soil, on its own terms. Europe holds the strongest hand at that table and has not played it. The AI Act is the most comprehensive AI safety regulation anywhere and it is in force. The Council of Europe’s Framework Convention is the first binding international treaty on AI. Yet having attended the UN’s first Global Dialogue on AI Governance in Geneva in July, (see our submission here), we didn’t see the EU driving the conversation. This means that:

• Europe should put senior, sustained representation into building a global AI governance body, and open the science-policy dialogues that make one work — starting with the infrastructure for that inside the EU, where researchers, officials and civil society can speak the same language..

But it’s not just AI. Now the public has had a taste of the heatwaves to come, the EU should get ahead of solar geoengineering governance before others set the terms. EU Foreign Affairs Ministers recognised in April that solar radiation modification has become geopolitical. Europe should take a leadership role that builds on their insights:

• Fund the research within an EU-coordinated agenda, bring in non-EU researchers and particularly those from the global South, and pair it with a firm agreement on non-deployment, as the Group of Chief Scientific Advisors recommended. 
• Start building international monitoring of the stratosphere now, because it will take years and the €10 million of Horizon funding is a first step rather than a plan. Our survey of European opinion in June found more nuance among citizens than the debate usually credits them with.

The thread running through all these recommendations isn’t really about any individual technology. Europe cannot assume it will control how a technology develops; AI has made that obvious. Nor can it assume the costs of getting it wrong land anywhere near the people who made them. This summer was a blunt enough lesson in who ends up carrying risk that has been deferred.

Sustained competitiveness requires foresight, resilience, and a clear-eyed view of what the future might hold. Europe’s advantage lies in its capacity to shape things early rather than react to them late, and its long history of building institutions that last through multiple generations. It’s a more nuanced play than simply cutting red tape, but it’s the right one for a continent that wants not only to prepare for – but also shape – its future.

If you are working on any of this and it would help to talk, get in touch. It is a small field, smaller than it should be, and we would rather be useful than right.