OpenAI adds a prominent AI-safety skeptic to its board committee with model-release authority
Paul Christiano -- RLHF co-inventor and former OpenAI alignment lead, now known for warning about loss-of-control risk -- joins the Foundation board's Safety and Security Committee, recusing himself from OpenAI-specific evaluations.
OpenAI announced on September 9, 2026 that Paul Christiano -- co-inventor of RLHF, founder of the Alignment Research Center, and OpenAI's own head of alignment from 2017 to 2021 -- is joining the OpenAI Foundation board and its Safety and Security Committee, chaired by Zico Kolter, along with a non-voting observer seat on OpenAI Group PBC's board. Christiano will recuse himself from OpenAI-related matters and model evaluations specifically.
A specific, on-record reason for the concern
Christiano's own stated reasoning for taking the seat, per his public statement: "public evidence from recent incidents suggests that this is not just a theoretical possibility" -- referring to loss-of-control risk as something increasingly grounded in observed behavior, not a purely hypothetical concern. That's a notably direct framing from someone now sitting on the committee with a voice, even without a formal evaluation role, in how OpenAI governs its own releases.
What changes for anyone evaluating OpenAI as a vendor
A named safety researcher known specifically for loss-of-control concerns now has a seat on the committee overseeing safety and security at the company building GPT-6 Astra and its successors. It doesn't change what the models can do, but it's a concrete governance signal worth weighing alongside any vendor risk assessment -- and a real possibility of future friction over release timing that's worth watching, not assuming either way.