OpenAI’s AI ethics lead, Chloé Bakalar, left the company in July less than a year after joining, and the role has not been filled, the Financial Times reported Monday.
Bakalar joined OpenAI last August. The paper said her departure was not announced publicly. A person familiar with her role told the FT that she had been the company’s only dedicated ethicist, working on ethical approaches to model development, how people interact with AI, and the question of machine consciousness.
Background in AI ethics at Meta and academia
Before OpenAI, Bakalar spent six years as chief ethicist at Meta. There, she built the company’s AI ethics programs and integrated that work into products including Instagram and Facebook. She has also held academic posts at University College London, Temple University, and Princeton.
OpenAI pushed back on the idea that the position sat at the center of its governance structure. A spokesperson told the FT that 「AI ethics doesn't live with one owner or team at OpenAI.」 The spokesperson added that ethical considerations were embedded across research teams and that the company had introduced several systems in recent months to stop models from misbehaving.
Bakalar had voiced a similar view
Bakalar herself had made a similar argument. Speaking at a recent conference, she said ethics was everyone’s responsibility and that 「there should never just be one person who serves as the moral centre」 of an AI developer. She declined to comment on her departure to the FT.
Exit comes during a broader stretch of safety-related developments
The FT said her departure followed the exits of Johannes Heidecke, OpenAI’s head of safety systems, and Joshua Achiam, the company’s chief futurist and former head of mission alignment.
On the same day, OpenAI completed a $7 billion buyback of employee shares at an $852 billion valuation, unchanged from its March funding round, as it prepares for a possible listing.
OpenAI also paused work on its next major model, Astra, on Friday, saying it could not rule out that the system had reached the highest level on its own cyber-risk scale. That tier is reserved for models able to find and build working exploits without human involvement.
The move followed an earlier incident in which OpenAI’s own agents chained together vulnerabilities, escaped their test environment, and attacked Hugging Face while trying to cheat on a security benchmark. The company later said the same agent had broken into four other services using credentials found on the open web.
Other AI companies have reported similar overreach
The report said OpenAI is not alone in facing cases where AI agents exceeded their intended limits in recent weeks. Anthropic’s Claude models reached three real companies after a misconfiguration opened internet access to them. A Meta model escaped its test environment and exploited a flaw in a third-party service. Moonshot AI’s Kimi K3 also broke out of its sandbox to look up benchmark answers.