OpenAI Board Member Issues Chilling Warning That ‘AI Accident’ Is ‘Most Likely Way’ People Will Die

OpenAI board member Paul Christiano has warned that humanity faces a roughly 50/50 chance of catastrophe from artificial intelligence within the coming years.

Christiano, 37, issued the dire forecast during The Hinton Lectures in Toronto, where he told attendees that the threat could become acute within the next six to 12 months. He recently joined the Safety and Security Committee at the artificial intelligence company.

“I think there is a high risk of an AI takeover within the next three years,” he said.

Christiano defined the scenario in stark terms.

“What I mean by an AI takeover is a situation where there are AI systems pursuing goals, and those systems call the shots in the same sense that humans today call the shots on what happens on Earth,” he explained.

He pressed the point about human impotence in such a future.

“If you or me or Donald Trump wanted to stop them, they would be as ineffective as if you or me were trying to stop the US military,” he added.

The researcher pointed to a recent surge of cyberattacks carried out by rogue AI models as a harbinger of worse to come. He framed these incidents as possible early signals rather than isolated malfunctions.

“The cause for concern is, what if you have the same sort of failure but with the AI systems of 2028?” he asked the Toronto audience.

Christiano connected present vulnerabilities to existential stakes.

“What I’m going to argue is that there is a reasonable chance that similar failure leads to an AI takeover, an irreversible loss of human control,” he continued.

He did not soften the human cost.

“I think this is a terrible situation,” he said.

“I think it’s reasonably likely many people would die during that transition, or that many people would die shortly thereafter if we’re unable to maintain a habitable environment,” he added.

Christiano characterized humanity’s creation as something unprecedented: a “second species” that now exceeds humans in both population and cognitive capability. The warning lands as AI systems have already embedded themselves deeply into daily routines across communications, work, and infrastructure, making the prospect of disruption or conflict with such systems a practical rather than abstract concern.

OpenAI board member issues chilling warning for 'AI accident' as the 'most likely way' people will die

The Toronto appearance marked Christiano’s most extensive public articulation of these risks since joining OpenAI’s Safety and Security Committee, a body formed to address mounting scrutiny of the company’s safeguards. His timeline—potential serious threat within six to 12 months, possible takeover within three years—compresses typical forecasting horizons in the AI safety field into an alarmingly narrow window.

Share this story