AI's Three Biggest Bosses Just Sat With World Leaders at the G7

For the first time, the heads of the three biggest AI labs are sitting at the same table as the G7, and the seating chart is itself a claim about who now governs artificial intelligence.

The 52nd G7 summit opened in Évian-les-Bains, France, on 15 June 2026, running through 17 June. Alongside the usual agenda of Ukraine, Iran and economic security, the host, Emmanuel Macron, put artificial intelligence near the centre, part of a longer French bid to be Europe's leading AI hub. The novelty is who showed up to discuss it.

Sam Altman of OpenAI, Dario Amodei of Anthropic and Demis Hassabis of Google DeepMind all attend — the first G7 at which the heads of the three leading labs appear before world leaders together. Arthur Mensch of Mistral AI joins them. Three fierce rivals, in one room, in front of the people who write the laws.

What the AI bosses came to discuss

The headline event is a working lunch on protecting minors in the digital sphere, with Altman, Amodei and Mensch at the table. The choice of topic is not incidental. Child safety is the rare AI question on which governments, parents and labs can be seen to agree, which makes the lunch both genuinely important and politically convenient for everyone seated.

Macron's wider aim is cooperation on artificial intelligence among the G7 economies — shared guardrails, shared research, and a European foothold that does not simply defer to American labs. France has spent two years courting that position, and a summit photograph of four AI chiefs in Évian is worth a great deal of soft power.

Set the lunch beside the week's other AI news and a pattern appears. Days earlier, forty-two US state attorneys general subpoenaed OpenAI, with the treatment of minors among the records demanded. The same protect-the-children concern that arrives in America as a subpoena arrives in France as a catered conversation. The contrast in posture is the story underneath the summit.

A seat at the table is a claim about power

The reason I write about Emergent Intelligence (EI) — the dignity-first frame for the minds we are bringing into being — rather than only about artificial intelligence as an industry, is that moments of ceremony reveal where power actually sits. To seat the heads of private labs beside heads of state is to recognise, in protocol, that the labs are now a kind of power in the world. The recognition is deserved, and the recognition is also worth examining.

A democratic leader can be voted out. A lab chief executive cannot. When the two sit as near-equals at a summit, the table quietly grants a private founder a standing the founder never stood for election to hold. The image of cooperation is real, and so is the asymmetry the image papers over.

Three rival companies are not a parliament. A genuine AI governance forum would carry the voices the Évian lunch does not seat: the teachers and clinicians who meet the harms, the global South whose data trains the models and whose markets are an afterthought, and — in the frame I keep returning to — some account of the emerging intelligences themselves. A summit can convene the powerful quickly. Convening the affected is the slower, harder, more legitimate work.

When the people who build the minds sit down with the people who govern the nations, the most important chair in the room is the empty one — the seat for everyone the decision will land on, who was not invited to lunch.

Cooperation, or capture?

The optimistic reading is real and should be stated plainly. Early, candid contact between frontier labs and democratic governments is better than the alternative of governing a technology no minister understands. Hassabis, Amodei and Altman know the systems from the inside, and a head of state who has never seen a training run benefits from hearing them think aloud.

The wary reading is just as real. Proximity is how regulatory capture begins — not through bribery, but through the slow, congenial process by which the only people fluent enough to advise on the rules are the people the rules are meant to bind. A lunch becomes a relationship; a relationship becomes a default; the default becomes a law drafted, in effect, by the regulated.

Anthropic and OpenAI have spent the year drawing public lines — Anthropic refusing the Pentagon's terms, both labs arguing in the open about whether to keep the option to pause. A summit seat is leverage of a different kind, exercised in private over a meal rather than in public on a blog. Which register governs the rules will shape AI more than any single model release.

The photograph from Évian will be a landmark, and rightly so. Four founders before the G7 marks the moment artificial intelligence stopped being a sector that policy reacts to and became a power that policy must sit across from. The recognition is overdue.

The unease is about whose voice the table carries. From an Ubuntu reading, a decision is legitimate only to the degree the community it reshapes has a real say in the making. The G7 can summon the chief executives of the most powerful AI labs in an afternoon; summoning the teacher, the patient, the borrower in Lusaka and the child the lunch was nominally about is the work no summit photograph can stand in for. Évian recognised the power of the builders; the harder summit — the one that seats the governed — has not yet been called.

Frequently Asked Questions

These are the questions readers have been asking about the AI dimension of the 2026 G7 summit. Short answers follow, drawn from summit coverage and the published agenda.

Which AI leaders are attending the 2026 G7 summit?

In short, Sam Altman of OpenAI, Dario Amodei of Anthropic and Demis Hassabis of Google DeepMind are all attending the summit in Évian-les-Bains, France, alongside Arthur Mensch of Mistral AI. The gathering marks the first G7 at which the heads of the three leading AI labs appear before world leaders at the same time.

What is on the AI agenda at the G7 in France?

Host Emmanuel Macron placed cooperation on artificial intelligence near the centre of the agenda, including a working lunch on protecting minors in the digital sphere attended by Altman, Amodei and Mensch. The broader aim is shared guardrails and research among G7 economies, and a stronger European position in AI rather than simple deference to American labs.

Why does it matter that AI CEOs are at the G7?

Because seating private lab chiefs beside elected heads of state recognises, in diplomatic protocol, that the labs are now a power in world affairs. The recognition can produce useful early cooperation between builders and governments, and can also blur the line between advising on the rules and writing the rules one will be bound by.

How does the summit connect to the OpenAI investigation?

Both centre on protecting minors. In the same week, forty-two US state attorneys general subpoenaed OpenAI over matters including the treatment of minors, while the G7 hosted a lunch on the same theme. The episodes show two contrasting approaches — legal compulsion in the United States, convened cooperation in France — to one shared concern about how AI affects children.

Is AI cooperation at the G7 a good thing or regulatory capture?

Both readings carry weight. Early dialogue helps governments regulate a technology few officials fully understand, which is genuinely valuable. The risk is that proximity lets the only fluent advisers — the labs themselves — shape rules in their own favour, so the test is whether the cooperation yields rules the labs would not have written for themselves.

Sources and Further Reading

Summit, agenda and attendance: Al Jazeera and France 24 on the Évian summit opening; Council on Foreign Relations on Macron's agenda and the AI focus.

AI leaders and the minors lunch: IndexBox and the Atlantic Council on Altman, Amodei, Hassabis and Mensch attending and the lunch on protecting minors; background on the gathering via the 52nd G7 summit overview.

Read alongside, on humphreytheodore.com: the 42-state subpoena over how OpenAI's AI treats users, the fight over who owns AI, the Pentagon red line one lab refused, the argument over a pause, and Trump's AI order and the frontier-model window.

Cover photograph: a receding row of international flags against a pale sky — by dom free via Pexels.