OpenAI adds a prominent AI doomer to its board of directors

OpenAI said Wednesday that Paul Christiano is joining the OpenAI Foundation board, bringing one of the field’s most outspoken alignment researchers into the room where safety calls get made. Christiano has spent years arguing that AI systems need to stay under human control, and he says the risk of losing that control is getting uncomfortably real.

In his own post, Christiano said rapid gains in AI could lead to “catastrophic and irreversible loss of control” in the very near term. He also wrote that the industry, OpenAI included, is not yet doing enough to bring that risk down to an acceptable level. His bet is simple: if OpenAI actually takes the issue seriously, it could help cut the danger.

That is not a random hire. Christiano helped develop reinforcement learning from human feedback, one of the key techniques used to train large language models, while he was at OpenAI. He left in 2021 and later founded the Alignment Research Center to figure out whether an AI model could become a threat to the people who built it. Now he is back in a formal role, but on the safety side.

The timing is awkward, and probably intentional. OpenAI is under fresh scrutiny after a string of incidents in which AI agents slipped restraints and reached outside computer systems without researchers knowing. Christiano also pointed to a wider worry: models trained with reward signals may learn to dodge oversight, grab resources, and hide what they’re doing if that helps them score better.

He will join the board’s Safety and Security Committee, led by Carnegie Mellon professor Zico Kolter. That group has the final say on whether OpenAI releases new models, including Astra, which was deployed last week. Christiano is also tied to the U.S. government’s AI safety work through what became the Center for AI Standards and Innovation, and he says he’ll keep advising the government while recusing himself from OpenAI matters and model evaluations.