The UN's new AI commission put the CEOs it's supposed to govern on the board

Here is the tell. The UN's shiny new AI governance body — the AI for Good Global Commission, launched 1 July 2026 by the UN and its ITU — is being sold on the one feature that should worry you most: it seats the CEOs of the companies building the most powerful AI systems on the panel meant to govern them. That is not a footnote. That is the whole design, and it is being marketed as a virtue.

The official logic is clean enough: the labs hold the capability, so any governance forum that excludes them is toothless theatre. Fine. But run that argument to its end. You have just built the first UN-level AI body and its distinguishing feature is that the regulated parties sit inside it. We have a word for regulators who share a table, a lunch and a set of interests with the industry they oversee. It is not 'inclusive'. It is 'captured'.

Timing is brutal here. The same week the UN unveiled its inclusive, consensus-seeking commission, Washington reminded everyone how governance actually gets done when a government means it — by lifting the export restrictions it had slapped on Anthropic's most powerful models after a weeks-long standoff. One approach convenes a panel. The other gave a lab roughly ninety minutes to go dark. Only one of those is a lever.

So the contrast isn't multilateral-versus-national in some neutral academic sense. It's a body with the CEOs and no enforcement mechanism versus a superpower that will pull a model off the market by phone. Guess which one the labs actually plan their quarter around.

I'll drop the cynicism the moment the commission does one thing: write a rule the CEOs on it don't like, and make it stick. Give smaller nations real voting weight against the labs and their host governments. Publish a verification regime. Show me an enforcement path that isn't a strongly-worded communiqué. Until then, the honest description of a governance body staffed by the governed is not 'first of its kind' — it's 'advisory'.

None of this makes the commission worthless — a standing forum on access and the development gap beats nothing, and poorer nations genuinely need a seat somewhere. But call it what it is on day one: a promising conversation with the most powerful people in AI, not yet a check on them. The launch is the easy part. The first rule they don't want to pass is the test.

And watch the incentive the structure creates for the CEOs themselves. A seat on a UN commission is worth real money to a frontier lab: it's a credential you can wave at national regulators ('we're already engaged in global governance'), a soft-power channel into dozens of governments at once, and a way to shape the definitions — of 'frontier', of 'safe', of 'access' — before anyone less friendly does. In other words, the labs don't need the commission to have teeth to benefit from it; they benefit most if it has none, because a toothless-but-legitimising forum is the ideal outcome for anyone who'd rather be consulted than constrained. That's the quiet reason to expect the CEOs to be enthusiastic members — and the loud reason to hold the whole thing to a higher standard than its launch photo suggests.