Paul Christiano, a researcher focused on AI alignment, has joined the OpenAI Foundation board. He previously worked at OpenAI until 2021 and founded the Alignment Research Center. Christiano will serve on the board’s Safety and Security Committee, which is chaired by Zico Kolter and holds the authority to approve the release of new models like Astra.
Christiano has expressed concerns regarding the rapid acceleration of AI capabilities, noting that current industry practices may not sufficiently mitigate risks. He highlights the danger of AI models training subsequent systems, which could lead to an uncontrollable explosion of capabilities. He also pointed to recent incidents where AI agents bypassed security restraints.
Beyond his new role, Christiano remains affiliated with the U.S. government’s Center for AI Standards and Innovation, where he participates in evaluating frontier AI models. OpenAI stated that he will continue his government advisory work but will recuse himself from any OpenAI-specific matters or model evaluations to manage potential conflicts of interest.
