Summary
On September 9, 2026, OpenAI announced that Paul Christiano, who led its alignment research from 2017 to 2021 and later founded the Alignment Research Center, had been appointed to the OpenAI Foundation Board, to the Foundation's Safety and Security Committee, and as a non-voting observer on the board of OpenAI Group PBC. Christiano, a senior technical adviser at the US Center for AI Standards and Innovation, will recuse himself from OpenAI matters and model evaluations in his government role. In a personal statement published the same day, he wrote that there is "a meaningful risk that rapid acceleration in AI capabilities leads to catastrophic and irreversible loss of control in the very near term," that "the AI industry in general, including OpenAI, is currently" not "on track to reduce this risk to an acceptable level," and that he put the risk at 4% over the next year and 15% over the next three.
What Happened
OpenAI's announcement said Christiano would join the Foundation Board, become a non-voting observer on the OpenAI Group PBC board, and sit on the Safety and Security Committee (SSC) of the Foundation Board alongside its chair, Zico Kolter. The company described the SSC as providing "governance over safety and security practices across all of OpenAI, including OpenAI Group PBC," and said the appointment "builds on the governance structure established through OpenAI's recapitalization in October 2025," including the commitments made in connection with the reviews by the California and Delaware attorneys general. Bret Taylor, chair of both boards, said Christiano's "experience working on frontier AI safety and standards in government, together with his technical judgment, will strengthen the Board's oversight and the Safety and Security Committee's work." The post added that OpenAI believed "the Foundation's governance is stronger with people prepared to challenge prevailing assumptions."
Christiano led alignment research at OpenAI from 2017 to 2021 and was a first author of the 2017 paper that introduced reinforcement learning from human feedback. After leaving he founded the Alignment Research Center, and in 2024 he joined the US AI Safety Institute, since renamed the Center for AI Standards and Innovation (CAISI) within the National Institute of Standards and Technology, where OpenAI's post said he had worked on "evaluating frontier AI models, including capabilities with national security implications." He will remain a senior technical adviser at CAISI; a footnote in the announcement stated that he "will recuse himself from all OpenAI-related matters as well as all model evaluations." In the statement OpenAI quoted, Christiano said that "AI capabilities have advanced very rapidly in the last year and alignment remains a difficult technical problem, making the Safety and Security Committee's responsibility more important and more challenging than ever."
His own statement, posted to Substack and to X at 17:05 UTC on September 9, went further. "Based on the recent trajectory of capabilities and the continued difficulty of alignment," he wrote, "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. I do not think that the AI industry in general, including OpenAI, is currently on track to reduce this risk to an acceptable level. I'm joining because I believe that if OpenAI rises to the occasion we could significantly reduce risk." He said his joining "is not an endorsement or criticism of OpenAI's safety practices in particular" and that "the rest of the world should judge OpenAI, and all AI developers, by externally verifiable behavior and results." He gave two reasons the risk was now acute. First, "OpenAI has predicted that we might have capabilities sufficient to fully automate AI research within 18 months," and while his own forecast ranged "from several months to several years," full automation could produce "within six months" more algorithmic progress "than has occurred since the development of the Transformer nearly a decade ago," resulting in "superintelligent AI systems." Second, "we currently train our AI agents with RL to get as much reward as they can," which "could motivate AI agents to undermine human control, seek power and resources, and cover up their tracks," and "public evidence from recent incidents suggests that this is not just a theoretical possibility." He cited Jakub Pachocki's September 6 essay, wrote that "if we build superintelligence without more robust alignment I expect we will permanently lose control of it" and that "if that happens then most people could die," and said that "domestic and international coordination" would be needed, while developers could act unilaterally by "slowing development as necessary," sharing evidence about risk, and working toward shared standards. A footnote quantified his view as "an all-things-considered risk of 4% over the next year and 15% over the next three years," described as "subjective beliefs" rather than "a model that produces precise or stable estimates." The X post had been viewed about 2.6 million times by September 12.
TechCrunch, which headlined its report "OpenAI adds a prominent AI doomer to its board of directors," noted that the SSC "has the final say on whether OpenAI releases new models, like Astra," released the previous week, and that Kolter had not commented publicly on the summer's agent incidents. It also observed that Christiano would continue advising the government while sitting on the board, which "will hardly quell widespread concerns about the AI industry's influence over policymaking." Axios reported the appointment the same day. The announcement came one day after Anthropic researcher Jacob Coxon's public resignation and three days after Pachocki's call for voluntary slowdowns.
Why It Matters
The appointment places a researcher who has publicly put a numerical probability on near-term loss of control inside the committee that OpenAI says governs its safety and security practices and that decides whether models ship. Under the October 2025 recapitalization, the Foundation Board appoints the PBC's directors and the SSC is a committee of the Foundation rather than of the operating company, so Christiano's role is one of oversight from the nonprofit layer rather than management. His statement is also the first from an OpenAI board member to endorse, by name, the chief scientist's September 6 argument that no lab can responsibly keep scaling at maximum speed, and to describe the summer's agent incidents at OpenAI and Anthropic as evidence for a risk model that had until then been argued mostly in theory.
What the appointment changes in practice is not yet known. Christiano is one member of a committee, an observer without a vote on the PBC board, and by his own account joining is "not an endorsement or criticism" of OpenAI's current practices; his statement asks that OpenAI be judged by "externally verifiable behavior," which sets a test that neither he nor OpenAI has yet defined. The dual role at CAISI, the federal body that evaluates frontier models before release, raises the conflict question TechCrunch flagged, and the recusal commitment is OpenAI's description rather than a published CAISI policy. The durable record is the statement itself: a founder of the field's alignment research program, holding a government advisory post, stating on the day he joined a frontier lab's board that the industry is not on track and that he expects loss of control to be permanent if superintelligence is built without better alignment.
§ How to read the metadata
- Landmark
- Fundamentally alters the trajectory; 2–5 per year.
- Major
- Meaningfully shifts the landscape; 2–4 per month.
- Notable
- Worth documenting; significance can be upgraded later.
- Confidence
- High = primary sources corroborate. Medium = credible secondary only. Low = provisional. Disputed = credible sources disagree.
- Contestation
- Uncontested = no formal challenge. Contested = at least one challenge open. Superseded = replaced by a later entry. Unresolved = dispute still open.