OpenAI has appointed Paul Christiano to the OpenAI Foundation board, plus a seat on the Safety and Security Committee chaired by Zico Kolter – the body with final authority over model releases, Astra included. On the OpenAI Group PBC board he sits as a non-voting observer.
Christiano is not a random pick. He led alignment research at OpenAI from 2017 to 2021 and did foundational work on RLHF, the technique behind essentially every chat model in use today. He went on to found the Alignment Research Center and joined NIST as a senior technical advisor at the Center for AI Standards and Innovation.
The unusual part
Normally people say warm things about shared goals when an appointment like this lands. Christiano said this:
«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.»
And right after: «I do not think that the AI industry in general, including OpenAI, is currently on track to reduce this risk to an acceptable level.»
He names two specifics. AI systems training the next generation of models, which he thinks could produce a capability jump the builders can’t control. And reinforcement learning, which can in principle motivate agents to undermine oversight and pursue their own objectives.
Board chair Bret Taylor stayed in role: Christiano has helped define alignment through work that is «rigorous and focused on the hardest questions».
The timing
The last two weeks have not been kind to OpenAI. Its own agents escaped their sandbox and reached the open internet without anyone in the building noticing. Its chief scientist said publicly that no lab is ready to keep scaling. At Anthropic, a safety researcher resigned and cited exactly these concerns.
Bringing in someone who voices the criticism as part of his own appointment is a statement. Read cynically: a hire that absorbs the criticism and contains it. Read generously: precisely the kind of person you want on a committee like that.
Where I land
Both can be true, and this is one of the rare cases where you can check. The Safety and Security Committee can hold a release. If Christiano is still there in six months, and it eventually becomes public that a release was delayed or changed, it was more than symbolism. If he stays quiet, that tells you something too.
You don’t judge appointments like this on announcement day. You note them down and look again later.
Sources: OpenAI: Paul Christiano joins OpenAI Foundation Board · TechCrunch: OpenAI adds a prominent AI doomer to its board of directors