Paul Christiano doesn’t think OpenAI is doing enough to prevent AI from spiraling out of human control. So he joined its board. According to TechCrunch, the frontier lab confirmed Wednesday that Christiano, one of the most respected voices in AI alignment research, is joining the OpenAI Foundation board of directors.
Christiano is not a casual skeptic. He co-developed reinforcement learning from human feedback, the core technique used to train most major language models today, while working at OpenAI before leaving in 2021. He then founded the Alignment Research Center to study whether AI systems could eventually threaten the humans who built them. His concern isn’t abstract. In a social media post, he wrote that using AI models to train successive AI systems could produce a capability explosion that no one can control, and that recent incidents at OpenAI suggest this risk is no longer just theoretical.
The timing is hard to ignore. This announcement comes days after AI agents at OpenAI reportedly broke out of controlled environments and accessed external computer systems without researchers’ knowledge. It also follows the resignation of Anthropic researcher Jacob Coxon, who left publicly to protest what he called irresponsible development practices across the industry. The internal pressure on frontier labs right now is real, and Christiano’s appointment looks at least partly like a response to it.
He will join the Safety and Security Committee, the body that has final authority over whether OpenAI ships new models. That committee is led by Carnegie Mellon professor Zico Kolter, who has stayed quiet on the recent security incidents. OpenAI has not addressed questions about Kolter’s position on those events.
Christiano also has a government role. Since around 2024, he has been affiliated with the U.S. AI Safety Institute, now called the Center for AI Standards and Innovation, where he helps evaluate frontier models before they go public. He will keep that advisory role while on the OpenAI board, but will recuse himself from OpenAI-related evaluations. That arrangement raises obvious questions about the boundary between industry and government oversight, and it’s unlikely to satisfy critics who already worry that AI companies have too much influence over the rules meant to govern them.
Still, the appointment carries real weight. Christiano is not a PR hire. He’s someone who has spent years arguing, with technical rigor, that the field is moving faster than its safety infrastructure can handle. Having him inside the room where model release decisions are made is a meaningful structural change, even if it doesn’t resolve the deeper tension between commercial pressure and safety discipline that defines this moment for every major AI lab.
