In a November 2026 interview, Geoffrey Hinton warned that an AI-related incident could disrupt electricity or financial systems later in 2026 or in 2027. His examples were an attack on electricity generation that causes a major brownout, or an incident that brings down the financial system for a week.
The scenarios behind Hinton’s warning
Hinton described both outcomes as possibilities. He used “Chornobyl” as a metaphor for a crisis that might force political leaders to pay attention. His forecast gives a broad time window, from later in 2026 through 2027, rather than a specific date.
Hinton’s proposal for reviewing AI models
Hinton proposed that companies provide strong statistical evidence to an independent authority that a model does more good than harm before release. He compared the idea to oversight of drugs by the U.S. Food and Drug Administration (FDA). The comparison describes the kind of independent review he wants; it is not an existing FDA approval process for AI models.
Hinton has also argued for regulatory review before AI products launch. NeoTeo’s earlier coverage of that proposal explains that earlier call.
How Hinton’s earlier concerns developed
In 2023 interviews with The New York Times and NPR, Hinton described a sharp change in his estimate of when AI might surpass human intelligence. He had once thought it could be 30 to 50 years away, or longer; that year, he said it might be as close as five years. These were his estimates at the time.
Those interviews also recorded concerns about false images, video and text, job displacement, autonomous weapons and the difficulty of controlling international competition. NPR covered his concern that sophisticated disinformation could interfere with elections. In a December 2025 CNN interview, Hinton said AI had progressed faster than he expected, particularly in reasoning and deception.
Hinton has said he believes catastrophe can still be avoided, but that doing so will take substantial effort.
