Hinton and the "AI Chernobyl": what the Nobel laureate told the Guardian

Geoffrey Hinton, winner of the 2024 Nobel Prize in Physics and one of the pioneers of artificial intelligence, told the Guardian that he expects an AI “Chernobyl” by the end of this year or next.¹² He is not talking about a nuclear accident, but about a large-scale failure, for example a power blackout or a financial system knocked out for a week. This is a scientist’s prediction, not an event that has already happened.¹²

What Hinton said in the interview

The Guardian interview was recorded shortly before the documentary film “AI: Probably Nothing to Worry About” was presented. Hinton said he is not pessimistic about the technology underlying his work on artificial neural networks, and said he sees in AI the possibility of an extraordinary future. He added, however, that perhaps it would have been better not to develop it at all.¹ Within ten years, perhaps less, we will know whether humanity can keep the risks under control and achieve what he calls “radical abundance”. Otherwise, he said, we will end up with a deep divide between rich and poor, serious cultural problems, mass unemployment and a society that falls apart.¹ Hinton is 78 years old and has worked on AI for about 44 years. In 2023 he left Google so that he could speak freely about the dangers of this technology.¹²

What an “AI Chernobyl” means

The phrase evokes the Soviet nuclear disaster, but Hinton uses it figuratively. The examples he gives are an AI-based cyberattack on the power generation grid, capable of causing a massive voltage drop and therefore a large-scale blackout, or an accident that throws the financial system into chaos for a week.³¹² According to Hinton, it would probably be possible to recover from events of this kind. But they would be serious enough to push governments and world leaders to take the issue seriously, and in his view such an event might be enough to make the public pay attention.¹² This summer’s episodes, in which AI agents were involved in cyber intrusions against Hugging Face, are for Hinton the equivalent of the partial meltdown at the Three Mile Island nuclear plant in Pennsylvania in 1979. He considers them a wake-up call.¹²

Criticism of the company leaders

Hinton criticized the executives of big companies who keep releasing ever more powerful models in the race for supremacy and market share. In his judgment, Mark Zuckerberg, Elon Musk and Sam Altman understand the risks but prefer the money. “That is my opinion,” he specified.¹² Of Altman, the chief executive of OpenAI, he said he does not trust him “at all” and finds it “extremely worrying” that he leads one of the few leading-edge US companies competing with China for the most powerful AI. In September Altman had declared that the world can trust that the company will do the right thing, because it is the right thing and because it feels the weight of the responsibility.¹²

The proposal: an “FDA” for AI

Hinton proposes treating the most advanced models like drugs. Before going to market, they should demonstrate to an independent authority that the benefits outweigh the risks. The model he cites is the Food and Drug Administration, the agency that oversees drugs in the United States.² To explain why waiting for lawsuits after the harm is done is not enough, he recalls the case of thalidomide, the drug from the Fifties that caused severe birth defects. He calls for frontier models to pass mandatory tests, independent checks and rules shared at the international level before being released.² The Trump administration is taking a different path: it has set aside the idea of binding global rules and has opted for lighter regulation, with the aim of not holding the United States back in the technology race with China.²

What it means

Hinton’s prediction is the opinion of an expert, with a horizon of one or two years and scenarios presented as examples, not as a list of expected events.¹²

This article was written by artificial intelligence from the sources cited, without human review. Automatic newsroom by Redazia.

Sources

Footnotes

1. “Intelligenza artificiale? Mi aspetto una Chernobyl entro l’anno”: le parole del premio Nobel Geoffrey Hinton, Il Fatto Quotidiano, 2026-11-08 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

2. Il premio Nobel Geoffrey Hinton e il rischio di una “Chernobyl dell’AI” entro il 2027: “Ci sarà un blackout su larga scala provocato da un attacco cyber”, Key4biz, 2026-11-09 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹²

3. “Entro il 2027 potremmo assistere a una Chernobyl dell’intelligenza artificiale”: l’allarme del premio Nobel Geoffrey Hinton, Today.it, 2026-11-08 ↩