Trump is right that AI is a race. But every race needs a track, and the Gulf could build it.
This week the President told the UN that whoever wins superintelligence wins, and that America rejects any global scheme of control. He's right on the first and half right on the second. Here's the half he's missing, and where it could be built.
On Tuesday 22 September the American President stood up at the UN and told the world where America stands on AI, with no holds barred. Whoever wins 'superintelligence' wins, he said (he has decided that "artificial" makes it sound fake, so superintelligence it is). The United States "totally rejects any attempt to construct a globalist scheme of control" for AI. And America does not retreat from a frontier.
Strip away the delivery and he was describing something real. The world's thinking on AI has split into two clean avenues: go fast, because otherwise you get left behind, or put the brakes on, because otherwise you risk losing everything. Washington has picked its avenue and is in no doubt about it.
And nowhere was that split clearer than on one single Wednesday earlier this month. On 2 September the Mayor of New York banned AI from every classroom in the city up to the age of 14 - nearly 600,000 kids - for a year while the effects are studied. On the very same day, the UAE made AI compulsory in every school in the country, public and private, from kindergarten up. One day, same question - completely opposite answers.
The safeguards exist. What's missing is anyone building them together.
I've been on both sides of this one. In my UK government days I was in the room for the conversations on AI and digital adoption policy, and I ran a city region whose whole mission was to digitally transform itself to work better and cost less (and I have the scars to prove it). I now live and work in Dubai, alongside a government that has decided AI isn't a policy area at all - it's the new operating system. So I'm seeing this first-hand, on a weekly basis. And I think both sides are missing the same thing. The President is right about the race. Where he's wrong, and where the Gulf comes in, is the assumption that how fast you run is the only way to "win" it.
Two very divergent paths
Western Europe has chosen caution: a very entrenched, very old-fashioned caution. The EU's AI Act is the perfect example, a rulebook so heavy that this summer Brussels had to delay its own flagship high-risk rules to 2027 and 2028, because nobody was ready for them. Europe wrote the rules before it built anything. Now it's delaying the rules and still not building anything.
America, the Middle East and, to be fair, China have gone the other way. Here the worry isn't what AI might do to people, it's being left behind. And the machinery of government runs accordingly. Since January the UAE's national AI system has sat as an advisory member of the Cabinet and of every federal board, and the target is agentic AI running 50% of government operations within two years. Nobody here is asking whether to adopt. They're embedding it deep into the Government and running as fast as the tech will allow.
The kids and the concrete
The physical manifestation of the digital rush tells the same story. In Abu Dhabi, the first 200MW phase of Stargate UAE - the foundation of the 5GW UAE-US AI campus - is well underway, barely 16 months after it was announced. In the UK meanwhile, data centres are queuing up to 15 years for a grid connection, and Whitehall's 'big ambition' is to cut planning consent from 18 months to 12. That sounds like progress until you remember that the Abu Dhabi campus went from announcement to poured foundations in less time than that. Forget enabling adoption; Britain can't even get the diggers in.
What that means in practice should worry any modern nation. Capital, talent and compute go where they're welcome and where they can actually be built. Western Europe is being left behind - on investment, on skills, on talent - because it's blocking every avenue.
Which brings me back to those two Wednesday decisions. As a parent with kids in a Dubai school, I know which one I'd want. Not because AI in a classroom is risk-free - it clearly isn't, and a 10-year-old with a chatbot doing their homework isn't learning anything - but because the UAE's approach is to educate kids for the world they actually live in, not the one we'd like them to live in. What AI is, how it gets things wrong, how to check it, and when not to use it at all. Banning it is putting our kids' heads in the sand. And if you don't expose them to something that's going to be a vital part of their lives, don't be surprised when it overtakes them rather than the other way round.
The bit nobody is doing
So do the fast movers have it right? On adoption, yes. But there's a bit missing, and it's missing everywhere. The safeguards exist. What's missing is anyone building them together.
It isn't that the Gulf is careless on safeguards - a lazy accusation from people who haven't looked. In June the UAE set up a new federal authority for AI and data, reporting straight to Cabinet, precisely to set the standards. It signed the Paris AI declaration when Washington wouldn't. The real point is different: every country, and every lab, is doing its safeguarding on its own. A model trained in one country, deployed in a second and used by a company in a third doesn't care whose rulebook it's under. Safeguards done in isolation aren't really safeguards, they're just gestures.
The reason we got here is a failure on both sides. The safety camp built its institutions - Bletchley, Seoul, the safety institutes - on the assumption that safety means slowing down. So the builders decided the whole thing was a brake and, at Paris, walked out of the room. We've ended up with a safety conversation with no builders in it, and a building race with no safety conversation in it.
If you want proof that the premise is still the problem, look at the last fortnight. Dario Amodei of Anthropic published an essay calling for the frontier labs to slow down together, and Sam Altman and Elon Musk both said he was right. Within two days the American President had dismissed the lot of it, and at the UN this week he made it official. The builders finally asked to do something together and, because the something was a brake, their own government said no.
So what?
It doesn't need to be a binary choice. Safeguards, yes. Getting left behind, no. The way to get both is a broad international coalition on safety, built from the start on the understanding that this isn't about clipping anyone's wings. The opposite of a scheme of control: unashamedly pro-adoption, with a remit that rules out telling any country or company how fast to go.
What it does instead is three things. Agree how far is too far: the handful of genuine red lines that every serious player already accepts in principle and nobody has written down together. Agree how safety and security get embedded as AI goes into government, critical infrastructure and defence, before a power grid goes down and we find out what was missing. And broad policy alignment, so a business deploying AI across 20 countries isn't navigating 20 rulebooks and quietly picking the loosest. Not one regulator. A common floor, and above it, a race.
Who builds the track?
Who convenes it? Not Brussels, which nobody building anything currently trusts. Not Washington alone, which nobody else will accept. The obvious candidate is the one region the fast movers trust because it's one of them, that has shown it will sign up to international commitments, that is building the institutions to make them stick, and that is one of the very few places Washington and Beijing will both actually turn up.
The Gulf could host this. Given how quickly things get built here, it probably should. The countries that work out how to keep a keen eye on safety, together, without leaving the next generation behind on what is probably a bigger moment of change than the industrial revolution (on that, at least, the President and I agree) - they'll own the next 20 years.
Sources
Remarks by the US President to the UN General Assembly, 22 September 2026, as reported by Fox News and IBTimes. NYC Mayor's Office, statement on the one-year moratorium on student-facing generative AI, 2 September 2026. Khaleej Times, UAE Cabinet approves AI curriculum for all public and private schools, 3 September 2026. UAE Cabinet announcements on the National AI System joining Cabinet and federal boards (June 2025) and the Federal Authority for AI and Data (June 2026). European Commission, Digital Omnibus on AI, in force 27 July 2026. Paris AI Action Summit declaration and signatories, February 2025. Barbour ABI, UK data centre pipeline analysis, August 2026; Department for Science, Innovation and Technology, AI Growth Zones policy paper. G42 and OpenAI, Stargate UAE announcement, May 2025. Axios, 12 September 2026, and NPR, 13 September 2026, on the Amodei essay, the responses from Altman and Musk, and the President's remarks.