Geoffrey Hinton warns EU leaders AI could outsmart humans within 18 months

Geoffrey Hinton has spent years telling anyone who would listen that AI will eventually get smarter than us. On Friday in Dublin, he brought a much shorter timeline to the most powerful room in Europe.

The Nobel laureate, often called the godfather of AI, was scheduled to brief the leaders of all 27 EU member states. His message: machines could surpass human intelligence within 18 months. That is a long way from the decade he used to cite.

A timeline that scares even Hinton

The 18-month window is not one Hinton invented on his own. It is the estimate he now hears from Paul Christiano, the AI safety researcher whom OpenAI added to its board in September.

Hinton has said that timeline frightens even him. Coming from the person who helped build the foundations of modern AI, that is not a small admission.

For context, here is how he described his own earlier position:

“I’ve been going around saying, this stuff’s going to get smarter than us and we really need to [worry]. It might happen within 10 years.”

The summit took place on November 13, 2026. It came just days after Christiano spoke at the Hinton Lectures, held November 9 to 11, 2026. At that event, Christiano raised alarms about a potential AI takeover within the next three years.

Hinton’s own estimate of the odds of a catastrophic AI outcome remains between 10% and 20%.

Why the Dublin summit matters

The gathering marked the first time EU leaders gave AI its own dedicated agenda item at a summit.

Discussions in Dublin also covered the EU’s AI Act. The law is designed to set a regulatory framework for deploying AI while trying to balance innovation against safety.

European Council President António Costa and European Commission President Ursula von der Leyen both acknowledged the need for stronger governance to manage AI risks. The discussions also floated the idea that Europe could help shape international standards for the technology.

Hinton’s warnings went beyond raw intelligence benchmarks. He has pointed to existential threats including loss of sovereignty and what he calls “runaway AI,” scenarios in which humans lose meaningful control over systems they built.

How Hinton got here

Hinton left Google in 2023, a move that freed him to speak openly about the dangers he saw in the technology he helped pioneer.

Since then, he has steadily shortened his estimates for when superintelligent systems might emerge. Christiano’s role adds another layer. He is a respected voice in AI safety research, and OpenAI’s decision to place him on its board in September gave his views more institutional weight.