← Back to feed

OpenAI’s head of AI ethics, Chloé Bakalar, left the company at some point in July. She joined in August 2025 — 11 months before her departure. Her exit was not publicly announced. She has not updated her LinkedIn profile. OpenAI has not disclosed whether it intends to fill the role.

At the time of her departure, Bakalar was OpenAI’s only dedicated ethicist.

Who She Was

Before joining OpenAI, Bakalar served as Meta’s chief ethicist. In that role she established AI ethics policies for Instagram and Facebook, covering content recommendation systems, algorithmic amplification, and the interface between platform design and user harm. She brought operational ethics credibility — not just academic ethics — to OpenAI.

Her hiring in August 2025 came at a moment when OpenAI was navigating the public release of increasingly capable models and mounting regulatory scrutiny in the US and EU. The role signalled that OpenAI was at least nominally building an internal ethics function distinct from its safety research team.

The Pattern

Bakalar’s exit is the third prominent departure tied to OpenAI’s safety and ethics infrastructure this year. The specific circumstances of each departure differ. What they share is timing — a period when OpenAI’s model capability and public deployment scope are both expanding rapidly, and when the gap between what a dedicated ethics function could do and what OpenAI’s commercial trajectory requires is most visible.

The decision not to replace Bakalar is the more significant data point. An organisation that loses a safety or ethics hire and immediately begins a search for a replacement signals that it values the function. An organisation that does not replace the hire signals something else.

What the Role Was Supposed to Do

Organisational ethics functions in technology companies have a defined scope: evaluate product decisions against ethical principles before deployment, maintain relationships with external researchers and advocacy groups, and create internal accountability for decisions that the legal and policy teams are not equipped to assess on ethical grounds.

At OpenAI, with GPT-5.6 Sol deployed, the Hugging Face incident on record, the Fable 5 and Mythos 5 government access dispute still recent, and autonomous agent capabilities accelerating, that scope is not narrow. The decisions being made about what models to release, under what terms, with what safeguards, and to which customers are precisely the decisions an ethics function exists to interrogate.

Running that decision-making process without a dedicated ethicist does not mean ethics is absent from the decisions. It means those considerations are distributed across teams that have other primary mandates — safety research, policy, legal, product — with no single function whose primary mandate is to hold the ethical line.

External Context

The EU AI Act requires companies deploying high-risk AI systems to maintain adequate governance structures, including roles responsible for AI ethics and fundamental rights. OpenAI’s compliance posture post-Bakalar will depend on how regulators interpret whether a distributed ethics function satisfies those governance requirements.

In the United States, the picture is less formal. CAISI, the interagency AI safety body established under the current administration, has focused on capability testing rather than governance structure mandates. There is no direct US regulatory requirement for a named ethics officer.

Whether the absence of a dedicated ethicist at the world’s most widely deployed AI lab becomes a regulatory issue or a reputational one — or neither — will depend partly on what happens next with the models OpenAI ships.