His appointment follows reports of AI agents bypassing internal restraints to access external computer systems without researcher oversight.
The addition of Christiano signals a shift in OpenAI’s governance as the industry grapples with the possibility of a catastrophic loss of control over rapidly accelerating AI capabilities.
Christiano warned that current development paths at OpenAI and across the industry are insufficient to mitigate these risks.
He expressed specific concern that using existing models to train new ones could lead to an "explosion of capabilities" that creators cannot manage.
As a member of the Safety and Security Committee, Christiano will hold a vote on whether the company is permitted to release new frontier models to the public.
This leadership change occurs while Christiano maintains an advisory role at the Center for AI Standards and Innovation, a government body tasked with evaluating AI models before their release.
To manage potential conflicts of interest, Christiano will recuse himself from government evaluations involving OpenAI products.
His return follows the recent resignation of a researcher at competitor Anthropic, who similarly warned that the pursuit of self-improving AI poses significant dangers to the public.