Paul Christiano, a board member of the OpenAI Foundation, told a Toronto audience that an AI accident is the most likely way most of them will die.
Christiano, who is also executive director of the Alignment Research Center, spoke at the opening lecture of a three-part series. "I think for most people in this room, the most likely reason they will die is because of an AI accident," he said. (Alignment research is the work of getting AI systems to pursue the goals their builders intend.)
His central worry is what he called an "AI takeover." He defined it as goal-pursuing AI systems that "call the shots in the same sense that humans today call the shots and what happens on Earth." In that world, he said, a human trying to stop them "would be as ineffective as if you or me was trying to stop the US military." He said such an outcome would likely mean mass death and perhaps environments humans cannot live in.
The passages quoted here carry no probability and no timeline. That makes this a judgment from an insider, not a measured forecast, and its weight comes mostly from who is speaking. Christiano sits on the board of the nonprofit tied to one of the labs building the systems he is warning about.
The warning itself has precedent. In 2023, lab leaders and researchers signed a one-sentence statement ranking AI extinction risk alongside pandemics and nuclear war. What is different here is the framing: a personal mortality claim aimed at a room of ordinary attendees. The two remaining lectures may supply the mechanism and the odds that the opening one, at least in these excerpts, does not.