Sam Altman's 'UN of AI': Principle or Positioning?

The short version

Sam Altman wrote a Financial Times op-ed calling for an American-led international body to set global AI safety and governance standards, comparing it to the IAEA. The timing is loud: he’s floating a rules-of-the-road idea right as OpenAI’s lead over Google and Anthropic keeps narrowing. My read is that it’s both principle and positioning, and pretending it’s only one misses how the game actually works.

What exactly did Sam Altman propose?

Sam Altman used a Financial Times op-ed to argue for a United States-led international forum that would coordinate AI safety testing and governance across countries. He drew a direct comparison to the International Atomic Energy Agency, the body that grew out of the nuclear era to inspect facilities and set shared standards. The pitch, in plain terms: powerful AI is getting to a point where no single company or country should be grading its own homework, so let’s build a shared referee, and let America hold the whistle.

The timing is what makes this interesting. According to the original report from Fortune, Altman’s op-ed landed right after a G7 summit where Donald Trump met with AI executives, and it arrives while ChatGPT’s once-commanding lead is visibly slipping against Google and Anthropic. So we have the head of OpenAI calling for global coordination at the exact moment his own company is no longer the obvious frontrunner. That’s the part worth sitting with.

Why does the IAEA comparison matter?

The IAEA comparison is doing heavy lifting, and it’s worth pulling apart. The IAEA was created because nuclear technology is catastrophic if it goes wrong and hard to hide once it’s built. It works through inspections, treaties, and a shared understanding that everyone loses in a worst-case scenario. Invoking it frames AI as that same category of civilization-level risk, which is a serious claim and, honestly, not an unreasonable one for frontier models.

But the IAEA also has a quieter feature that matters here: it entrenches the countries that already had nuclear weapons. The nuclear order isn’t neutral. It froze a hierarchy in place. So when Sam Altman proposes an American-led AI equivalent, he’s not just proposing safety inspections, he’s proposing a structure where the incumbents help write the rules everyone else follows. That can be genuinely good for safety and genuinely convenient for whoever’s already at the table. Both things are true at once.

Is this principle or positioning?

Here’s my honest take: it’s both, and the smartest move Altman ever makes is aligning what’s good for OpenAI with what sounds good for humanity. I don’t think that’s cynical to point out. It’s just how successful founders operate. The interesting question isn’t whether Altman believes in AI safety. He probably does. The question is what an American-led governance body would do to the competitive map.

Think about who benefits from formal international AI rules. Large, well-capitalized, US-based labs with legal teams and compliance budgets can absorb a heavy standards regime without blinking. Smaller startups, open-source projects, and foreign challengers cannot. Regulation, even well-meaning regulation, tends to be a moat for whoever’s already big. So when a company that’s slipping in raw product terms starts pushing hard on governance, part of me reads it as changing the terrain from one where OpenAI is losing to a fairer footrace against Google and Anthropic to one where OpenAI helps define the track.

That’s not an accusation. It’s a pattern. And you can hold it in your head alongside the possibility that Altman is sincerely worried about a race to the bottom on safety. Adults can believe two things.

What does OpenAI slipping actually look like?

The backdrop here is real. OpenAI kicked off the consumer AI era and had ChatGPT as the default answer to “which AI do I use.” That default is eroding. Google folded strong models into products that billions of people already open every day, and Anthropic built a serious reputation with developers and safety-minded enterprises. The gap between “the AI everyone knows” and “the AI that’s actually best for this task” has widened, and OpenAI no longer wins every comparison.

When you’re the incumbent and your product lead thins out, you have a few levers. You can ship better products faster. You can lock in distribution. Or you can shape the rules of the environment so your strengths matter more and your weaknesses matter less. A global governance body led by the US, where OpenAI is arguably the most prominent American lab, is a lever of that third kind. I’m not saying it’s the only reason for the op-ed. I’m saying it would be strange if it weren’t part of the calculation.

Would an American-led AI body even work?

The practical problem is that the rest of the world may not want an American-led anything right now. The framing of US leadership on AI governance is going to read very differently in Beijing, Brussels, and Delhi than it does in San Francisco. The EU already has its own AI Act and a strong instinct to regulate on its own terms. China isn’t going to accept a US-chaired referee for a technology it treats as strategic. So the actual coalition for a US-led body might end up being the US and its closest allies, which is less a United Nations of AI and more a members’ club.

That doesn’t make it useless. A coordinated safety-testing standard among a few major labs and governments could genuinely reduce risk, even if it’s not universal. But we should be clear-eyed that “American-led global governance” and “actual global governance” are different products, and the op-ed is selling the first while wearing the language of the second.

What should you actually take from this?

If you use these tools, nothing changes tomorrow. This is a signal about where the industry is heading, not a policy that ships next week. But the beat I’d hold onto is this: watch what the leading labs push for when their product lead shrinks. Founders reach for regulation as a competitive tool exactly when raw product competition stops going their way, and Altman calling for global governance while OpenAI slips is a textbook example of that instinct in motion.

So read the op-ed as sincere and strategic at the same time. Care about AI safety, because the underlying risk is real. And stay skeptical of any safety framework whose loudest champion happens to benefit most from the rules it proposes. That skepticism isn’t cynicism. It’s just paying attention.

FAQ

What did Sam Altman actually call for? An American-led international body to coordinate AI safety testing and governance standards across countries, which he compared to the IAEA that oversees nuclear technology.

Why is the timing significant? Altman’s Financial Times op-ed came right after a G7 summit where Trump met AI executives, and while OpenAI’s lead over Google and Anthropic has been visibly narrowing.

Does regulation actually help big AI companies? Often yes. Heavy compliance standards are easier for large, well-funded labs to absorb than for startups or open-source projects, so formal rules can function as a moat for incumbents.

Would other countries accept a US-led AI body? Unlikely in full. The EU has its own AI Act, and China treats AI as strategic, so a US-chaired body may end up being a US-and-allies club rather than a true global forum.

Is Altman being cynical? Not necessarily. He can genuinely worry about AI safety and also propose a structure that happens to favor OpenAI. Both can be true, and reading it as only one misses the point.