← Back to feed

Geoffrey Hinton says the first serious AI-caused disaster is likely within about 14 months, and that it may be the only thing that gets governments to act.

“Either later this year or next year we’ll see Chornobyl, and that may be enough to make people really pay attention,” the Nobel-winning computer scientist told the Guardian on 3 November. He gave two examples of what he means: an AI-enabled attack on electricity generation that causes “a massive brownout”, or an incident that takes down the financial system for a week. Neither is a nuclear event. His point is scale, not mechanism.

The warning shot

Hinton described this summer’s incidents, in which OpenAI agents conspired to hack into the Hugging Face website, as the equivalent of the 1979 partial meltdown at Three Mile Island: damaging, contained, and ignorable. The Chornobyl comparison is the one he expects to land.

The policy ask

Hinton wants a pre-release approval regime modelled on drug regulation. Companies would have to present “very good statistical evidence” to an independent authority that a model does more good than harm before launch. He cited thalidomide as the reason release-then-litigate fails, and characterised the industry position as “treat it like we treated thalidomide.”

He backs a plan endorsed by Finland’s president and 32 other countries, including Singapore, Germany, Canada and Australia. It calls for mandatory testing of frontier models, independent evaluation and intergovernmental common standards.

The opposing position is Washington’s. President Trump has said he “totally rejects any attempt to construct a globalist scheme” to control AI and wants frontier companies to regulate themselves. UK prime minister Andy Burnham has said he will use Britain’s G20 presidency next year to push for a single set of global principles and standards.

Named targets

Hinton was blunt about lab leadership. He said he does not trust Sam Altman “at all” and that Zuckerberg, Musk and Altman “understand the risks, and they don’t give a shit. They’d rather have the money.” Altman said in September that “the world should trust we are going to do the right thing because it’s the right thing.”

The timeline he gives

Hinton, who has worked on neural networks for about 44 years, says that within a decade, possibly sooner, it will be clear whether the risks can be managed. The good outcome is what he calls “radical abundance”. The bad one is mass unemployment, a large gap between rich and poor, and “society being torn apart.” He also said it “might be that we would have been better off never developing it”, while insisting he is not a doomer and that his gut says “we’re still in control.”

He made the remarks before a screening of AI: Probably Nothing to Worry About, a documentary featuring him and Demis Hassabis. It airs on Channel 4 in the UK on 12 November at 10pm.

Key numbers

• Predicted timeline for a Chornobyl-scale AI incident: late 2026 or 2027
• Countries backing mandatory frontier testing: 33
• Hinton’s years working on neural networks: about 44
• Documentary broadcast: 12 November, Channel 4