Emrys Schoemaker, Executive Director, Global Digital Governance Lab, Global Governance Centre, Graduate Institute; Senior Associate for Digital Policy and Advisory, Caribou
Keywords: AI governance, Politics of inevitability, Global Majority, Sovereignty through collaboration
The case for radical action on AI is often built on an underlying assumption, and that assumption is about to be put to most of the world: that a single, contestable forecast about the technology’s trajectory is already settled, and that on the strength of that certainty, ordinary rules are a luxury and hard bargains the only responsible course. Frederike Kaltheuner has set out how this works in her response to Europe 2031, the much-discussed scenario in which the continent, having failed to secure sovereign access to frontier AI, slides into economic irrelevance and is left with three unpalatable options: become an American protectorate, turn to China, or fade. The scenario is a powerful one vivid and justifies the anxiety. But its sharpest flaw, as Kaltheuner shows, is the frame itself: it treats the market’s evolution as a force of nature to which Europe can only submit on the best terms available, when in fact that market is still being shaped, and perhaps the more important question is by whom.
That diagnosis has broader application than the essay that prompted it. The politics of inevitability is not being sold to Europe alone. Across a single fortnight this July it was put to everyone at once, and the audience least able to refuse it is the one that most repays attention: the middle powers and the Global Majority, the states unlikely ever to host a frontier lab and now being asked, in effect, to choose someone else’s approach to governing AI
Four governance models in a fortnight
Consider what unfolded between the first week of July and the third.
On 6 July, in Geneva, the UN convened the first Global Dialogue on AI Governance, and António Guterres told the plenary that every nation needs a seat at the table. On 14 July, Demis Hassabis, CEO of DeepMind, published a manifesto calling for a US-anchored, industry-funded standards body modelled on FINRA (Financial Industry Regulatory Authority) – a private AI regulator, financed by the labs, answerable to Washington, that would vet frontier models before they reach the American market. On 17 July, in Shanghai, Xi Jinping personally opened the World Artificial Intelligence Conference and unveiled a new China-led cooperation body pitched explicitly to the Global South. And alongside all of this, running since last December, sits Pax Silica: the US-led coalition binding chips, compute and critical minerals into a single alliance whose membership is conditional on geopolitical alignment, the arrangement that already decides who may buy the compute on which any market runs, and with which Hassabis’s Washington-facing standards body seems, broadly, to align.
These are not versions of a single exchange but a contest between four models of governance – arguably three, given that Pax Silica and Hassabis’ call align – coupling a grip on the supply chain with an industry proposal to write the rules for the market that runs on it. This sets out access as conditional on political alignment, with the terms set in Washington. The Chinese approach arrives in the language of inclusion, courting the Global South with a more equitable-sounding Chinese approach than the Western summits have extended. The UN argues for something neither bloc is offering, a multistakeholder model accountable to all states rather than to one capital or one industry. however, for a capacity-constrained state, the question each poses is the same: which governance approach do I align with, and whose architecture do I legitimate through alignment? That is how the politics of inevitability works at global scale, not by presenting a single offer to accept or refuse, but by narrowing the field to a menu of blocs and letting the narrowing imply that choosing between them is the whole of the available agency.
A starker choice
If this is the trap set for Europe, it closes harder on the Global Majority. Europe holds cards: ASML1 and a genuine chokepoint in lithography, regulatory weight scaled to its market, capital and industrial depth, and the option, however arduous, of building alternatives. It is already trying to exercise them, through the EuroStack campaign and the Commission’s Tech Sovereignty Package, insufficient, on any honest reading, to the scale of the dependency they address, but instruments nonetheless. A capacity-constrained state in the Global Majority has fewer, and the choice it confronts is correspondingly starker: not “submit to dependence or build sovereign capacity,” but “accept this patron or that one.” The particular danger of the politics of inevitability is that it closes off the very question such a state most needs to keep open: whether a third option exists at all.
Whether a third option exists
That question turns on more than the political economy of blocs. It turns on a reading of the technology itself, particularly whether the frontier’s lead is durable or fragile, whether capability stays concentrated or commoditises, whether open-weight models remain good enough to build on. If you treat the frontier’s dominance as permanent and there is no option but to pick a patron; read the same evidence as a still-moving target, with cheaper open models and diffusing expertise, and there is room to build. The choice of forum follows a judgment about the technology, and the politics of inevitability works by fixing that judgment in advance, so that a claim about where the technology is heading quietly forecloses the political options before anyone examines them.
This is why sovereignty through collaboration, the regional and plurilateral pooling as the practical form autonomy takes, is not wishful thinking but a live strategy, and one already taking shape. In November 2025 the Smart Africa alliance established an Africa AI Council to pool resources across member states, building on the $60 billion Africa AI Fund announced at Kigali earlier that year, an attempt to negotiate as a single digital marketplace rather than as fifty-odd states approaching the frontier one at a time. The model practitioners invoke is CERN: shared regional infrastructure no single state could justify alone. If the only choice on offer is which bloc to join, the task is to build what is not on offer, pooled capacity that constitutes a bargaining position, collective procurement that generates demand for open and multilingual models, shared compute at continental scale. None of this is easy or assured; the Africa AI Fund remains far more announced than disbursed. But it is the difference between shaping the market and being allotted a place in it.
The same sentence, spoken twice
The sharpest image of this time was not a launch but a repetition, a single line delivered from two stages eleven days apart. In Geneva on 6 July, Guterres said every nation needs a seat at the table. In Shanghai on 17 July, standing beside Xi Jinping as the WAIC body was announced, he said it again, almost word for word: a technology that will shape the future of humanity must be shaped by all of humanity, not governed by a handful of countries or companies. The same claim, pressed onto each successive stage as the working architecture of AI governance assembled around it. The UN was not ceding the ground; it was staking a claim to it wherever the summit happened to be.
To read this as hypocrisy would be easy but mistaken. Universality is the one asset the UN uniquely holds, so universality is what it carries into every room, though a principle asserted from stage after stage is not yet a governance capability. Guterres could name the divide precisely, a third of humanity still offline, compute and investment concentrated in a few states and firms, and naming a divide is not closing it. And yet the UN remains the one forum outside the emerging US–China split, the only venue where a state need not declare a bloc before it takes part. The Security Council’s permanent five can each stall collective action; but the wider multilateral system belongs to no single power, and that is the one place a middle power can act without first choosing a patron.
That points to what the UN is actually for. Its advantage is not speed, industry is always quicker; nor compute, which the blocs will always hold more of. It is legitimacy in the service of a path that belongs to no bloc: convening the states that need pooled capacity, helping them arrive at a shared strategy, lending the weight of the multilateral system to arrangements no one of them could stand up alone. The most consequential thing Guterres did in Shanghai was not to repeat the seat-at-the-table line but to arrive with more than principle, pointing to a nascent Global Network for AI Capacity Building and forthcoming recommendations on a Global Fund for AI. Whatever becomes of any particular vehicle, that is work the multilateral system, through agencies that have spent years building capacity on the ground, is already equipped to deliver. Convening is what the UN does; enabling durable, independent capacity is closer to the UN’s development purpose.
Whether it can is the question worth studying rather than assuming, and the one the Global Digital Governance Lab is being built to pursue, from inside International Geneva, where the convening happens, but with the analytical distance to ask what convening can and cannot deliver. Under what conditions does regional pooling yield real leverage rather than shared dependence? When does collective procurement shift what the frontier is prepared to offer? The answers will decide whether the next phase of all this, the capacity networks, the funds, the regional arrangements now forming, becomes a genuine third path or another well-meant principle nobody was equipped to act on.
The politics of inevitability works by persuading people the future is already decided. On the evidence of the last few weeks, it is not. It is being decided now, across these forums and others still to come. The question that matters is whether the states with the least leverage are equipped to help shape it, or asked only to choose the terms on which they submit.
- The Dutch multinational that is the world’s exclusive supplier of the extreme ultraviolet (EUV) lithography machines required to manufacture advanced semiconductors used in AI and high-performance computing ↩︎