Elon Musk, Sam Altman, Dario Amodei, Demis Hassabis and Mark Zuckerberg. Most people would struggle to name all five, yet this small group, wealthier than most nations, is setting the direction of the most powerful technology ever built.
Stuart Russell, a computer science professor at UC Berkeley, puts it simply. The rest of us were never asked. There was no vote, no referendum, no moment where the public was consulted on whether this should happen or who should be trusted to do it.
The strangest part is that the people building AI agree it is dangerous. In May 2023, the leaders of the major labs signed a one-sentence statement placing the risk of human extinction from AI alongside pandemics and nuclear war.
Amodei has put the chance of something going catastrophically wrong as high as 25%. Everyone agrees it could end very badly, and they are racing one another to build it anyway.
Normally a race with stakes this high sits with government, as it did with the Manhattan Project or Apollo. This time it is private enterprise, and the question is not whether these people are good or bad. It is who gave them the right to decide, and what the rest of us can do about it.
The promise that AI would level things out
For a long time the story was that AI could close gaps rather than widen them. It was cheap to access and easy to use, and early studies found it often helped the least experienced workers the most, pulling weaker performers closer to the strongest.
Ethan Mollick of the Wharton School makes this point in his book Co-Intelligence. Microsoft AI's Mustafa Suleyman has framed the technology as a potential leveller too, though not without risks.
The CEO of OpenAI no longer seems to believe it. In a 2026 interview with Nicholas Thompson, Altman said he now worried about a world where access stayed limited and the existing rich bid up the price, deepening the divide.
He has cooled on universal basic income and moved on to talking about collective ownership of compute instead. The optimistic story assumed AI would spread like past technologies. The economics of how it is built point the other way.
Upgrade to Absolutely Agentic Premium
Join our growing subscriber community and get access to:
Our daily Agentic Intelligence newsletter curates all the best stories from the top AI newsletters, development labs, news websites and social media into one email.
Ad-free podcast - premium ad free RSS feed that you can put into your podcast app of choice. We publish about 3 hours of video a month and now its available in audio form.
Subscriber area includes long form guides to the most important AI trends affecting content and marketing, along with automations and prompts that will save you a lot of time.
Why it matters who holds the power
If the technology is going to be held by a few people, it matters a great deal who those people are. The closer you look, the harder it becomes to feel reassured.
In April 2026, the New Yorker published an investigation into Altman built on interviews with more than a hundred colleagues. At its centre were internal memos from former chief scientist Ilya Sutskever, and the first word on the list was lying.
Musk is more straightforward, because he says what he thinks. In early 2026 he predicted AI would soon exceed all human intelligence, and that humans would not stay in charge. His best case for us was the way we treat chimpanzees: we could wipe them out, and have chosen not to.
Amodei of Anthropic occupies the moral high ground more convincingly than most. Even so, his warnings of a white collar jobs bloodbath travel fast precisely because controversy travels fast, which makes them harder to take at face value.
Google may be best placed to win, with its scale and its own chips. But its 90% grip on search and its shift to AI generated results are taking a wrecking ball to the wider internet.
A race with no referee
If the people cannot restrain themselves, the obvious answer is that government should. At the end of May, the White House prepared an executive order setting up a voluntary review of the most powerful models before release.
Hours before signing, the President postponed it. His reasoning was that it got in the way of leading China. The administration's AI Action Plan, published in July 2025, lists removing red tape ahead of anything about safety.
The contradiction ought to stop us. When the United States last believed a technology was this consequential, it built it itself. The atomic bomb was not produced by a private company optimising for shareholder returns.
There are fair counter-arguments. AI is a general purpose technology, not a bomb, and concentration may be the price of building it carefully. But those are arguments about outcomes. The objection was always about consent.
The Musk Altman case did not rule that OpenAI kept faith with its founding mission. It ruled that the question came too late. Whether a non-profit for all of humanity became an engine for the enrichment of a few was left undecided, and elections across the West may be the first fought over who controls this technology and on whose behalf.
