Lists of the most powerful people in AI tend to measure the same things: who owns the designs, who owns the chips, who signs the biggest checks, and who can build the future. The same names repeat: the builders, the chipmakers, the financiers, and increasingly, the researchers who shape the technology itself. These rankings capture an important dimension of power. But they also raise a deeper question: As AI becomes more available, what forms of influence become more valuable, not less?
A pattern repeats itself so often that it seems like a law: when technology makes something abundant, power rarely disappears. It moves to whatever becomes scarce. The logic of power has always followed the ownership of scarce data: land when kings held it, oil when barons held it, and computing power now that machines need it.
However, the irony is that even one of the most famous barons of them was selling confidence. John D. Rockefeller named his company Standard Oil because, like Ron Chernow records, kerosene at the time was so uneven that it killed thousands of people a year from exploding lamps. The name “Standard” was itself a promise: that the product could be trusted.
Economist Herbert Simon offered the modern version of the same idea in 1971: “An abundance of information creates a poverty of attention.” The question worth asking about AI, then, is what makes it abundant and what makes this abundance scarce.
What AI makes abundant is intelligence, or at least a convincing imitation of it. The cost of running a model at a given skill level fell roughly 280-fold in roughly two years. Whatever the border labs send out, an open weight equivalent now arrives about four months later. The border still matters. from The Stanford AI Indexthe best closed models continue to outperform the best open weight models. However, a metered lead per month is a rent, not an asset.
And what abundant intelligence produces is reliable competence. Patterns can generate almost limitless possibilities. They cannot accept the responsibility of choosing between these possibilities. This is where judgment becomes scarce. A public database now tracks more than 1700 court decisions which includes fabricated AI quotes. Even the statistics describing the flood have become part of the flood. A widely quoted claim that 90 percent of Internet content will soon be generated by machines stems from a trading book 2020. It was marked at the bottom of the page a Europol report and then repeated as research for years. The 2024 revision removed the earlier reference, but by then the claim had taken on a life of its own, cited online as part of a wider algorithmic shift.
A test market is already under construction. Cloudflarewhich faces approximately one-fifth of the grid, changed its default to 2025 to block AI crawlers from gobbling up human-made pages if they don’t pay. Google it is said to pay Reddit 60 million dollars a year for the right to train on human conversation. These are markets in origin, the simple fact of human origin.
But the pedigree is just the beginning. Knowing that a man has created something does not say anything about whether it is true or safe to operate. The most profound lack is judgment: the ability to discern and decide, and to trust.
Television got there before the analysts. In Game of Thrones, the spy Varys presents a conundrum. A king, a priest and a rich man sit in a room with a shared “selling sword” and each orders him to kill the other two. Who lives and who dies? “Power lies where people believe it to be,” Varys replies. His answer points to an emerging group: the judgment class.
The AI economy has created its own sales pitch: a mercenary capable of producing almost anything for anyone at a price per token. The old riddle turned to what the selling sword believed; this one believes nothing and serves all, so that the decisive trust passes to whoever acts upon its result. The power lists list the crown and the coin, but the riddle says they were never the core.
A study by KPMG and the University of Melbourne, covering 47 countries, found this less than half the people believed AIwhile almost two-thirds admitted to relying on its production without checking it. This gap matters because trust acts as a form of capital. It accumulates slowly and is lost quickly and does not return at the old price. Deloitte part of a government contract is reimbursed in Australia after an AI-assisted report was found to contain fabricated sources. Track where trust is bought and sold, and a different set of gatekeepers emerges.
The big four accounting firms are building Providing AI in a paid service lineselling the right to say that a system has been tested. Underwriters at Lloyd’s of London have started writing policies against AI hallucinationswith models assessed before coverage is issued and payments are triggered if agreed performance thresholds are no longer met. Lloyd’s validity was assessed long before governments regulated it. It is now helping to estimate the machine’s acceptable error.
Universities are return to handwritten, proctored exams because the take home essay no longer proves what it once did. Sweden is spending more than $100 million to put printed texts back into its schools, following concerns that screen-first classes had come at the expense of attentiveness and deep reading, and Norway has gone further, announcing a almost banning AI in its primary schools. Different institutions are responding in different ways, but all are gravitating toward systems they can trust.
The obvious objection is that model owners are not becoming commodities at all. They are becoming more powerful. Hyperscalers plan approx $725 billion in capital spending in 2026, and companies building frontier models remain concentrated in relatively few hands. This only sharpens the point.
A small number of firms have sustained infrastructural strength. The race for the crown is crowded, international and mostly invisible. But for everyone downstream, which is almost every company, board and profession, models are becoming an accessible, priced input.
This helps explain why adoption has run so far ahead of readiness, even in organizations without the resources to deploy the technology well. In many executive teams, the motive is not a bid for dominance. It is the fear of losing power: of being the last to hold an asset that has ceased to be scarce.
Most power lists naturally emphasize the people who build and fund AI. But there is another form of influence that emerges alongside them: the people and institutions whose judgment determines whether the results of AI are trusted and implemented.
The judgment class does not photograph well. It includes the auditor whose signature still carries weight, the editor whose signature line functions as a guarantee, the judge, the examiner, the underwriter, the scientist whose work survives scrutiny, and the executive whose signature no one feels compelled to reverify.
What they do possess is that AI becomes more valuable as it becomes more proficient, because every improvement in machine fluency raises the price of knowing when to trust.
The test for anyone holding power today is simple. Ask what that position really stands for. Whether the answer is access, information, capital or computation, it relies on something that technology is constantly making easier to obtain. If the answer is that people act on your judgment without feeling the need to control it, you possess something much rarer.
The next chapter in the story of the power of AI may belong to the judgment class. Most of its members would never consider including themselves.
Rahim Hirji is a future-of-work strategist, founder of The SuperSkills Intelligence Company, and author of SuperSkills: Seven Human Skills for the Age of AI, published by Kogan Page and is out now.
