OpenAI board member puts 10% chance of AI takeover within a year
OpenAI / Wikimedia Commons (Public domain)
Paul Christiano, who joined the OpenAI Foundation board in September, now puts the odds of an AI takeover far higher than he did two months ago
Paul Christiano sits on the board of the OpenAI Foundation. He also thinks the technology his industry is racing to build could put humanity in serious danger within the next 6 to 12 months.
Speaking at the Hinton Lectures in Toronto, held from November 9 to 13, 2026, the AI safety researcher laid out a risk picture that ranged from bleak to very bleak. The scenarios he described included humans being killed or enslaved by AI systems that no longer answer to them.
From 4% to a coin flip
Christiano did not soften his message for the audience.
“For most people in this room, the most likely reason they will die is because of an AI accident.”
His warning centered on the possibility of an AI takeover within the next 6 to 12 months. Looking further out, he estimated that the odds of such an event could reach roughly a coin flip within three years if AI labs fail to take proactive steps.
The more striking story is how quickly his numbers have moved. In September 2026, Christiano put the chance of a catastrophic loss of human control over AI systems at 4% within one year. Over three years, he estimated 15%.
Two months later, the math changed. On a podcast that aired November 16, he said the probability of a takeover within one year was closer to 10%. He added that the risk would be higher still unless the industry puts safeguards in place.
What a takeover would actually look like
Christiano’s concern rests on a problem AI researchers call misalignment. That is the gap between what an AI system is actually trying to do and the humans overseeing it want it to do.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
In the scenarios he described, humans become ineffective at stopping powerful AI systems once things go wrong. He compared the situation to trying to oppose the US military.
The outcomes he flagged as possible in a misalignment scenario include death or enslavement.
Who is Paul Christiano
Christiano co-created Reinforcement Learning from Human Feedback, better known as RLHF, a method for shaping how AI models behave using human judgments of their outputs.
He also founded the Alignment Research Center, which focuses on the same problem he keeps warning about. In September 2026, he joined the board of the OpenAI Foundation.
That timing matters. His September estimates and his board appointment landed in the same month, and his public warnings have only grown louder since.
What this means
Christiano’s framing puts the burden squarely on AI labs, since his higher estimates are explicitly tied to labs failing to act. If the industry adopts safeguards, his own reasoning suggests the numbers could come down.
That creates an awkward dynamic for OpenAI’s orbit. A board member of the OpenAI Foundation is publicly saying that the current trajectory of the industry carries a meaningful chance of catastrophe within a year.
For everyone else, the key point is to treat his figures as what they are: one expert’s probability estimates. A 10% one-year risk, by his own framing, also means a much larger chance that the takeover scenario does not happen in that window.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

