OpenAI announced on September 9 that Paul Christiano is joining the OpenAI Foundation board and its newly formed Safety and Security Committee. He will also sit in on OpenAI Group PBC board meetings as a non-voting observer. The committee is chaired by Carnegie Mellon University professor Zico Kolter, and OpenAI's own description of its mandate stops at one word: governance — overseeing safety and security practices across all OpenAI entities, including OpenAI Group PBC.
His last stint at OpenAI ended before 2021
Christiano led OpenAI's alignment research from 2017 to 2021. He was among the foundational contributors to reinforcement learning from human feedback (RLHF), the method that later became standard practice for training large language models. After leaving OpenAI in 2021, he founded the Alignment Research Center (ARC), a nonprofit focused on aligning advanced AI systems with human interests. Since 2024, he has served as a senior technical advisor at the U.S. Center for AI Standards and Innovation (CAISI), formerly the U.S. AI Safety Institute.
Under the arrangement announced this week, his CAISI role requires him to recuse himself from all matters related to OpenAI, as well as all model evaluation work.
His own verdict
"I now believe there is a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control in the very near term."
The sentence that follows is that he does not believe the AI industry, OpenAI included, is currently on a path to bring that risk down to an acceptable level. That is what a newly appointed board member wrote in his own appointment announcement.
Foundation board chair Bret Taylor's assessment was that Christiano's work is rigorous and speaks directly to the hardest questions raised by increasingly capable systems.
The second alignment researcher on a ten-person board
Laid out together, the newly announced Foundation board reads: chair Bret Taylor, members Adam D'Angelo, Sue Desmond-Hellmann, Zico Kolter, Paul M. Nakasone, Adebayo Ogunlesi, Nicole Seligman, David Vélez and Robin Vince, plus CEO Sam Altman.
Roughly speaking, the backgrounds break down like this: a majority come from product and corporate governance; Nakasone comes from the national security establishment as a former director of the National Security Agency; two or three seats have investment and finance backgrounds; and until now, Kolter was the only member whose primary work was alignment research. With Christiano, that number becomes two.
Going from one to two is the entire change that can be verified on paper. Whether it translates into actual decision-making weight depends on what authority the Safety and Security Committee actually holds — and the announcement doesn't address that directly. Whether it can require delaying a model release, veto a deployment, or what conditions would trigger either move: none of that appears in this announcement.
A board shaped by the restructuring
This appointment comes after OpenAI completed its corporate restructuring. Under the plan announced in October 2025, the Foundation holds roughly 26% of OpenAI Group PBC, valued at about $130 billion at the time; Microsoft holds 27%; employees and other investors hold the remaining 47% combined. That makes the Foundation board simultaneously the governing body of a charity and the source of board seats for the company's largest single shareholder.
TechCrunch's report on the appointment cited two pieces of background: a recent incident in which an AI agent broke past its constraints and accessed external systems without researchers' knowledge, and the resignation of Anthropic researcher Jacob Coxon over safety concerns. The same report also raised a lingering question — that even with recusal clauses in place, the industry's influence over policymaking remains an open issue.
For now, outside observers can only verify the list itself. Whether this committee leaves a public record around the next model release is something that will only become clear when that moment arrives.
Sources: OpenAI official announcement, CocoLoop, TechCrunch, Unite.AI; the board list, committee mandate and recusal terms are verified against the official announcement, and post-restructuring equity stakes and valuation are based on publicly disclosed figures.