The Sovereign Middle

I have been thinking about a question that doesn't get asked enough. When people talk about AI governance, the conversation almost always reduces to a binary. Washington versus Beijing. The American model versus the Chinese model. The Brussels Effect maybe gets a nod, but mostly as a kind of third-party referee. And then everyone else, the other roughly 190 countries that will actually have to live with whatever comes out of this contest, is treated as scenery. I think that framing is wrong. Not just morally, although it is that too. I think it is empirically wrong. And the more time I spend reading actual UN documents and watching what is happening in Geneva and Abu Dhabi and Seoul and Brasília, the more I suspect we are missing the real story. The real story is the sovereign middle. What I mean by that By "sovereign middle" I mean a specific group of states that have three things at once. They have enough technical capacity to be taken seriously in any room. They have enough diplomatic neutrality to be trusted by both sides. And they have enough at stake, economically and strategically, to actually do the work rather than make speeches about doing the work. If you list the countries that fit that description today you get something like this. The UAE, Singapore, Saudi Arabia, South Korea, Finland, Estonia, El Salvador, Brazil, Morocco, Switzerland, Japan, Norway, the Netherlands. Not a coalition. Not a bloc. Not even a recognised category in international law. But quietly, almost without anyone naming it, this group is doing more concrete work on AI governance than either of the two superpowers it sits between. Let me give you the evidence. The Geneva clue On 6 and 7 July 2026, the first Global Dialogue on AI Governance will convene in Geneva. It is the first formal United Nations mechanism for ongoing intergovernmental conversation about how artificial intelligence should be governed. The Pact for the Future established it. Resolution A/RES/79/325 made it real. It is, by any measure, the most important multilateral AI venue ever created. Look at who chairs it. Egriselda López, the Permanent Representative of El Salvador. Rein Tammsaar, the Permanent Representative of Estonia. A Central American country of six million people. A Baltic country of one and a half million. Neither one a superpower. Neither one even a recognisable middle power in the traditional sense. And yet they were appointed by the President of the General Assembly to co-chair the most consequential AI governance dialogue in the world. That choice was not random. It was a signal. The UN is saying, in the only language it really speaks, that this conversation cannot be led by the players who built the technology. It has to be led by the players who can listen. What the middle is actually doing Estonia has been quietly running one of the most digitised states on earth for two decades. They have lived through Russian cyberattacks, they have built x-road, they have a level of practical experience with digital governance that most G7 countries can only theorise about. When Estonia speaks about AI, it speaks with the credibility of having actually done it. El Salvador is the unexpected one. Most people remember it for the Bitcoin experiment and not always charitably. But what gets missed is that the country has been actively cultivating a reputation as a small-state convenor on digital matters. Hosting the AI dialogue chair is part of a deliberate strategy. Small states with nothing to lose can do things mid-sized states with too much to lose cannot. Then there is the UAE, which is the case I know best. The UAE published its National AI Strategy in 2017, the first country in the region to do so, and updated it as the Strategy 2031, with a target of AI contributing AED 335 billion to GDP by the end of the decade (UAE Ministry of Cabinet Affairs, 2018). It established the world's first Ministry of State for Artificial Intelligence in October 2017. In May 2025 it signed the US-UAE AI Acceleration Partnership, and shortly after announced Stargate UAE, a one gigawatt compute cluster sitting inside a five gigawatt UAE-US AI Campus in Abu Dhabi, with G42 as the regional anchor and OpenAI, Oracle, NVIDIA, Cisco and SoftBank as partners (G42 press release, 22 May 2025). The first 200 megawatts is scheduled to come online in 2026. This is not symbolic infrastructure. This is the largest single AI compute deployment outside the United States. And it sits inside a country of fewer than ten million people, with a foreign policy doctrine that has been quite deliberately, and quite publicly, built around being the convenor that everyone can talk to. Why the superpowers can't lead this There is a structural reason for all of this, and once you see it, it is hard to unsee. The United States has the technology. It has the companies. It has, by a wide margin, the largest concentration of frontier AI capability in human history. But it cannot lead a global governance process for a simple reason. It is the country being governed. Any rule it agrees to constrains its own industry. Any rule it pushes is suspected of serving its own industry. There is no version of this conversation where the US is a neutral broker. It is the incumbent, and the world knows it. China has a similar problem from the opposite direction. Its model of AI governance is genuinely sophisticated. Some of the most thoughtful regulatory thinking on generative AI has come out of Beijing in the last two years. But the political distance between Chinese governance and the rest of the world is too wide to bridge in a single multilateral forum. Trust takes decades to build. China is starting that work, but it is starting late. This is why, when the UN actually had to design the architecture, the architects ended up being middle powers. Not because of some idealistic commitment to fairness. Because of practical necessity. If you want a conversation that 193 countries can actually join, you cannot start it with a country that scares half the room. The UAE's specific position I want to be careful here because I am Emirati and I am writing about my own country. Let me state the case as I would state it about anyone else. The UAE has a very specific advantage in this moment that almost no one else has. It is one of the few states that has serious technical capacity in AI, serious sovereign compute, serious investment vehicles like MGX, and at the same time a foreign policy that maintains working relationships with both the United States and China at a level of trust that very few others can match. This is not an accident. It was built deliberately, over twenty years, by people who understood that smaller states cannot afford ideology in foreign policy. They can only afford usefulness. And in the AI governance debate, usefulness means being able to convene. The October 2025 UAE statement to the Third Committee of the UN General Assembly explicitly mentioned the UN Youth Delegate Programme and the country's focus on "innovation, such as space and Artificial Intelligence" (UAE Permanent Mission to the UN, 2025). That is the kind of sentence that sounds boilerplate until you realise it is positioning. The UAE is telling the world that its youth-facing AI work is part of its foreign policy, not a side project. The 2026 UN Water Conference, which the UAE will co-host with Senegal, is the same logic applied to a different file. Co-host with a Global South partner. Position yourself as the bridge. What the Independent Scientific Panel tells us There is one more data point worth examining. In February 2026, the UN General Assembly appointed forty members to the Independent International Scientific Panel on Artificial Intelligence. They were selected from more than 2,600 applicants across 140 countries. No more than two members from any single nationality. Each region of the world represented. The panel reports annually to the Geneva dialogue. The vote was 117 in favour to 2 against. The two against were Paraguay and the United States. Read that again. The United States, the country with the deepest AI capability in the world, voted against the establishment of a scientific panel on AI. Its representative cited sovereignty concerns. Whatever you think of the merits, the diplomatic signal is unmistakeable. The dominant AI power has just publicly declined to lend its full credibility to the most important scientific body the UN has ever established on this technology. That is the gap the sovereign middle is going to fill. Whether it wants to or not. What this means in practice I do not think the sovereign middle will write a treaty. I do not even think it will produce a coherent doctrine. What I think it will do, over the next five years, is something more interesting and more durable. It will set the tempo. Through the Geneva dialogue. Through the Scientific Panel reports. Through working groups that meet quietly between sessions. Through bilateral arrangements like the US-UAE AI Acceleration Partnership that turn principles into practice. Through the slow, almost invisible accumulation of habits and precedents that, two decades from now, people will look back on and call "international law." This is how multilateral systems actually work. Not through dramatic moments but through unglamorous middle states that keep showing up, keep convening, keep submitting written inputs, keep co-hosting side events when nobody else will. A closing thought There is a temptation, when writing about the Gulf and AI, to either celebrate or warn. Either this is the future and we should all defer, or this is dangerous and we should all worry. I don't find either of those framings useful. What I find useful is the recognition that something genuinely new is being attempted. Not by Abu Dhabi alone. Not by Tallinn or San Salvador alone. By a quiet, distributed network of middle powers that have realised, before most observers did, that the AI governance vacuum is real and that someone has to fill it. If they get it right, the next decade of AI rules will not look like the trade rules of the twentieth century, written by a few large states for everyone else. They will look like something stranger and more democratic. Rules written by the countries that had the least to lose, the most to learn, and the most reason to listen. I find that hopeful. And rare. References G42. (2025, May 22). Global Tech Alliance Launches Stargate UAE. Retrieved from https://www.g42.ai/resources/news/global-tech-alliance-launches-stargate-uae UAE Ministry of Cabinet Affairs. (2018). UAE National Strategy for Artificial Intelligence 2031. Retrieved from https://ai.gov.ae United Nations. (2024, September 22). Pact for the Future, including the Global Digital Compact. A/RES/79/1. United Nations General Assembly. (2025, August 26). Modalities Resolution on the Global Dialogue on AI Governance and the Independent International Scientific Panel on AI. A/RES/79/325. United Nations. (2026, February 12). Statement by the Secretary-General upon the appointment of the members of the Independent International Scientific Panel on AI. Retrieved from https://press.un.org/en/2026/sgsm23016.doc.htm United Nations. (2026). Global Dialogue on AI Governance: Frequently Asked Questions. Retrieved from https://www.un.org/global-dialogue-ai-governance/en/faq CSIS. (2026). The United Arab Emirates' AI Ambitions. Retrieved from https://www.csis.org/analysis/united-arab-emirates-ai-ambitions UAE Permanent Mission to the United Nations. (2025, October). Statement to the Third Committee of the UN General Assembly. New York.