Paul Christiano joins the OpenAI board, warning of loss-of-control risk

OpenAI announced that alignment researcher Paul Christiano would join its Foundation board as a non-voting observer and sit on the Safety and Security Committee, the body chaired by Zico Kolter that oversees safety and security practices across OpenAI. Christiano co-invented reinforcement learning from human feedback, led OpenAI’s alignment research from 2017 to 2021, founded the Alignment Research Center, and is a senior adviser at the US Center for AI Standards and Innovation (CAISI) — the role through which he said he would recuse himself from OpenAI-related evaluation work.

In an unusually blunt personal statement, Christiano said the appointment was neither an endorsement nor a criticism of OpenAI’s current safety practices, but that he had accepted because he now saw an acute near-term danger:

If we build superintelligence without more robust alignment I expect we will permanently lose control of it. If that happens then most people could die.

— Paul Christiano, personal statement

He wrote that full automation of AI research could trigger a rapid intelligence explosion — potentially “more algorithmic progress” within six months than since the invention of the Transformer — and that he did not think the industry, OpenAI included, was on track to reduce the resulting risk to an acceptable level. He said he had resonated with chief scientist Jakub Pachocki’s “An Alien Mind” essay, and argued the world should judge AI developers “by externally verifiable behavior and results.” OpenAI’s announcement tied the appointment to the governance overhaul agreed with the California and Delaware attorneys general during its 2025 recapitalisation. The move added to a fortnight of senior figures voicing loss-of-control concerns, alongside Amodei’s call to pace the frontier.