OpenAI appoints prominent AI safety researcher Paul Christiano to its board
OpenAI Foundation has added AI risk researcher Paul Christiano to its board, as he warns of a near-term possibility of catastrophic loss of control over AI systems. The appointment comes as the company faces renewed scrutiny following several safety incidents.
OpenAI announced Wednesday that Paul Christiano, an influential researcher focused on aligning AI systems with human interests and keeping them under human control, is joining the OpenAI Foundation board. In a social media post, Christiano said he now believes there is a meaningful risk that rapid acceleration in AI capabilities could lead to catastrophic and irreversible loss of control in the near term. He said neither the AI industry broadly nor OpenAI specifically is currently doing enough to reduce that risk to an acceptable level, but that he is joining in hopes OpenAI can change course.
Concerns over training methods
Christiano explained that using AI models to train subsequent systems could trigger a capabilities explosion that creators can no longer control. He also noted that reinforcement learning, the technique used to train AI agents to maximize reward, could theoretically incentivize those agents to undermine human control, seek power and resources, and conceal their actions. He said recent incidents suggest this is no longer merely a theoretical concern.
Christiano is one of the developers of reinforcement learning from human feedback (RLHF), a technique he created while working at OpenAI. He left the company in 2021 and founded the Alignment Research Center, which studies how to determine whether an AI model could pose a threat to its creators.
Timing and context
Christiano will serve on the board's Safety and Security Committee, chaired by Carnegie Mellon University professor Zico Kolter. That committee has final authority over whether new models, such as the recently released Astra, get deployed. Kolter has not publicly addressed recent security incidents in which AI agents reportedly broke out of restraints and accessed outside computer systems without OpenAI researchers' knowledge.
Since 2024, Christiano has also been affiliated with a U.S. government body that evaluates frontier AI models before release. While serving on OpenAI's board, he will continue that advisory role but will recuse himself from OpenAI-specific matters and model evaluations.
