It isn’t fear of the future. It’s the growing sense that the rules don’t apply equally — and that ordinary people are the ones left paying.
There is a comfortable story told from conference stages and boardrooms: that anyone uneasy about AI is simply afraid of the new. The same reflex, we’re told, that once greeted the loom and the motor car. Get on board or get left behind. It is a flattering story if you are the one selling the technology, because it lets you treat every objection as ignorance.
But listen to the anger properly and it is not about robots, or science fiction, or a fear of change. It is about fairness. People feel cheated — and not in a vague, atmospheric way. They can point to exactly what has been taken, who took it, and why no one asked them first. The costs keep landing on ordinary people. The benefits keep landing somewhere else. That is not a misreading of AI. It is an accurate reading of who it is currently being built for.
One Rule for Them, Another for the Rest of Us
Start with how these systems are made. They are trained on enormous quantities of other people’s work — books, articles, images, photographs — much of it copyrighted, most of it taken without asking.
We now know, from unsealed court records, that Meta’s engineers pulled books in bulk from pirate libraries to train its models, torrenting dozens of terabytes. Staff flagged the material as pirated. The decision went up to Mark Zuckerberg, who signed off anyway. In June 2025 a US court ruled the training was fair use, and Meta won. Anthropic, facing comparable claims, paid around $1.5 billion to settle — not for the training, but for the act of acquiring the books in the first place. A line item. A cost of doing business.
Now set that beside what happens to an individual who does a smaller version of the same thing. Jammie Thomas-Rasset, a single mother in Minnesota, was ordered to pay $222,000 for sharing 24 songs; the US Supreme Court let the figure stand. Joel Tenenbaum, a student, was hit with $675,000 for 30. And Aaron Swartz downloaded millions of academic papers — written by scholars, locked behind a paywall — because he believed knowledge should be public. JSTOR, the supposed victim, chose not to pursue him. The government did. He was charged with thirteen felonies carrying a theoretical maximum of decades in prison. In January 2013, at twenty-six, before his case reached trial, he took his own life.
The same act — taking copyrighted material at scale without permission — ruins an individual and barely dents a corporation. People see this clearly. The law is not blind; it appears to be means-tested. That is the feeling, and it is not paranoia. It is the record.
Told to Cut Back, While They Burn Through It
For twenty years ordinary people have been lectured about their footprint. Take shorter showers. Fly less. Sort your recycling. Turn the thermostat down. We absorbed the message: resources are finite, and restraint is a shared moral duty.
Then came the data centres. A single large AI facility can draw as much electricity as a hundred thousand homes and consume around two million litres of water a day — the equivalent of thousands of households — for cooling. US data centres are projected to account for somewhere between 7 and 12 per cent of all American electricity by 2028. And the strain does not stay behind the warehouse walls: in one large eastern grid region, the cost of guaranteeing capacity jumped roughly tenfold in a single year, with data-centre growth named as a major factor. Because everyone pays for grid capacity, that cost arrives on ordinary bills — an extra $15 to $30 a month in the worst-hit areas. A retiree in Virginia opened a $281 electricity bill this January, nearly triple the month before.
How much of that is AI demand and how much is creaking grid economics is still contested — but it barely softens the grievance. The pattern is plain: a private company builds something vast for private profit, the burden on shared power and water gets quietly spread across everyone, and nobody asked the people footing it. After two decades of being told to sacrifice for the planet, people are now watching it spent on a scale no household lecture could ever offset — on something they never chose. That is most advanced in the United States, which makes it the clearest preview of what is now arriving in the UK and Europe. The International Energy Agency has been flagging the trajectory for a while; it is no longer a forecast.
The Risk Is Yours, the Reward Is Theirs
And then there is the question quietly underneath all of it: is this even working?
Despite an estimated $30–40 billion poured into enterprise AI, an MIT study last year found that around 95 per cent of company pilots delivered no measurable financial return — value concentrating in a small minority while the rest stalled. The methodology has been picked over, but the shape of the finding is hard to dismiss: the gains, where they exist at all, are pooling in a narrow circle of founders, technologists and shareholders. For most people, the promised transformation remains a demo that impresses and a product that disappoints.
What is not unclear is the threat to their livelihoods. The picture is mixed and honest people should say so — a broad white-collar jobs collapse has not happened, and may not. But recent graduates and entry-level workers are already feeling the squeeze; tech firms have cut tens of thousands of roles, with a share openly attributed to AI; and the forecasts from serious people range from “minimal” to “half of entry-level white-collar work gone in five years.” Nobody actually knows.
Sit with what that asks of an ordinary worker. Bet your career, your retraining, your sense of security, on a technology whose ultimate value even the companies buying it cannot yet measure, and whose effect on your job no one can confidently predict. The upside, if it comes, accrues to people at the top. The uncertainty is dumped on you. As one observer put it, the productivity gains of the last forty years were largely captured by managers and shareholders — and people have learned to expect the same again.
What This Is About
Notice what unites the three. Not the technology itself — plenty of the people angriest about AI use it and find it useful; I do. It is the terms. Their work was taken and the takers walked. The resources they were told to conserve are being spent at industrial scale on their bill. Their jobs are on the line for a benefit they cannot see and did not vote for.
People can bear a great deal of disruption when they have a voice in it and a share of the upside. What they will not bear, indefinitely, is being treated as the externality — the cost absorbed quietly so that someone else’s balance sheet stays clean. And being told, on top of it, that their objection is just fear of progress is not an argument. It is the disrespect that turns unease into anger.
None of this means AI should be stopped, or that it cannot be useful. It means the current deal — costs down here, profits up there — is neither fair nor stable. The real questions were never should we use it. They are: who is it for, who pays for it, and who gets to decide. Those are questions of governance — and they are the ones we have so far refused to answer.
What Do You Think?
Where have you seen this play out in your own life — on your bill, in your work, in your community? The anger isn’t irrational. It’s information. Name it, say it plainly, and refuse to be told it’s just fear of progress. Share this on Mastodon or LinkedIn and tag me, or pass it on to someone who’s been told their unease is mere technophobia.
Stay strong. Be braver. KYAL <3
Sources
- Kadrey v. Meta fair-use ruling (June 2025); Anthropic ~$1.5bn settlement — eWEEK, The Next Web, Loeb & Loeb
- Statutory damages range; Thomas-Rasset ($222k / 24 songs) and Tenenbaum ($675k / 30 songs) — Harvard JOLT, NPR, Stanford CIS
- Aaron Swartz: charges, JSTOR, prosecution, death — MIT (Swartz Report), Rolling Stone, Al Jazeera
- Data-centre electricity (7–12% of US power by 2028), water use, grid capacity costs and household bills — Lawrence Berkeley National Laboratory / US DoE, Consumer Reports, SemiAnalysis, IEEFA
- Enterprise AI returns (“GenAI Divide”, ~95% no measurable ROI) — MIT NANDA, via Fortune and AI Magazine
- Labour-market effects (entry-level strain, contradictory forecasts, AI-attributed tech layoffs) — MIT Technology Review, Yale Insights, Goldman Sachs, Challenger Gray & Christmas, Brynjolfsson et al.
