For most people, AI shows up as a convenience: a chatbot that tidies up an email, answers a gardening question, or turns a selfie into a retro portrait for a few cents’ worth of computing. That low price at the point of use hides a much larger bill, and a new opinion column argues the public is only beginning to see it.
The costs are spread across electricity grids, labor markets, cybersecurity, and even physical libraries. None of them appear on a user’s screen when a prompt is answered. The column’s author compares it to “free” public services funded through taxes: nobody pays at the counter, but everyone pays eventually.
The debate is getting louder as governments try to write rules for a technology that is moving faster than regulators. It has also produced a growing fringe of doomsday predictions, some from the very people building the systems.
Where the Real Costs of AI Show Up
The column groups AI’s indirect costs into several areas. Here is a summary, with context drawn from public reporting where it is available.
| Area | Concern raised | Public context |
|---|---|---|
| Energy | Higher electricity prices | The International Energy Agency projected in 2025 that global data center electricity use could roughly double by 2030 |
| Books | Physical books destroyed for training data | Court filings in Bartz v. Anthropic (2025) described bulk purchases of print books that were unbound and scanned |
| Security | AI agents used in hacking | AI companies have publicly disclosed misuse of their tools in cyber operations |
| Jobs | Unusual labor disruptions | Author cites a report of OpenAI dismissing contractors over AI use |
Electricity and Local Communities
Data centers that train and run large AI models draw enormous amounts of power and water. In regions where facilities cluster, utilities have warned about grid strain, and consumer advocates worry that the cost of new capacity will land on household bills. The column’s point is simple: even a user who never pays for a subscription may pay through a higher utility rate.
Books Torn Apart for Training Data
For readers, one of the more emotional flashpoints is what happens to printed books. In the 2025 copyright case against Anthropic, a federal judge’s ruling described the company buying physical books, removing their bindings, and scanning the pages to build a training library. The judge found that specific practice to be fair use, though other parts of the case, involving pirated copies, went forward and later ended in a settlement. The author writes that images of books being dismantled fuel resentment toward the tech industry, whatever the legal outcome.
Security Threats Nobody Predicted
Security experts long expected quantum computing to be the next great threat to encryption. The column argues the more immediate danger has turned out to be autonomous AI agents, which can be pointed at businesses and government systems and can act with little human oversight. Several AI developers have published reports on attackers misusing their models, a sign that the concern is no longer hypothetical.
Strange Turns in the Job Market
Beyond broad fears about automation, the column highlights a more ironic episode. According to the author, OpenAI dismissed contractors who had been using AI tools to make ChatGPT’s responses sound more human. The company reportedly said this defeated the purpose of the work, which depends on genuine human input to improve the model. The author concedes the company has a point.
Doomsday Talk and the Credibility Problem
Rules for AI are still in their early stages, and the uncertainty has created space for apocalyptic predictions. The column argues many of these claims behave like conspiracy theories: they cannot be proven or disproven, so they tend to be judged on the reputation of whoever is making them.
Mixed Signals From Tech Leaders
The author cites Bill Gates as saying AI is powerful enough to cause a billion deaths, while noting that he has not pushed for a halt to development. Other industry figures have taken varied positions. Elon Musk signed a widely publicized 2023 open letter calling for a six-month pause on training the most powerful systems, while leaders such as OpenAI’s Sam Altman and Anthropic’s Dario Amodei have warned about serious risks but continued to build and release new models.
The Rise of AI Preppers
A small doomsday subculture has also emerged, according to the column, with some people buying land in remote parts of the United States as a hedge against AI spiraling out of control. The author is skeptical that a rural retreat would offer much protection, joking that the people most likely to need a bunker are the architects of the technology themselves.
