OpenAI's only ethicist exits after less than a year
OpenAI lost its only dedicated ethicist — the third senior safety figure out this year — while a North Korean hacking unit quietly assembled a private AI stack, and the lab began hiring like an energy trading desk.
Chloé Bakalar, OpenAI's AI ethics lead, has left the company less than a year after joining, the Financial Times reports — leaving the lab without a dedicated ethicist and no named successor. Bakalar came over from Meta in August 2025, where she had been chief ethicist since 2019; her exit was never publicly announced and was not reflected on her LinkedIn. She follows safety systems head Johannes Heidecke and chief futurist Joshua Achiam, both of whom announced departures in July, making her the third senior safety figure out this year — part of a wave that has seen around half of OpenAI's safety researchers leave since 2023-24, per press tallies. OpenAI says ethics work is embedded across model teams rather than concentrated in one role, but the pattern leaves a live question for anyone relying on internal review to catch problems before models ship: who owns those calls now, and what does a lab's safety exodus signal to enterprise buyers in regulated industries?
North Korea-linked hackers are running AI models on their own hardware instead of renting intelligence from Western clouds, according to a report from South Korean cybersecurity firm Genians. The group, Kimsuky, set up local environments with Ollama, GPT4All and Msty plus retrieval-augmented search, alongside agent frameworks, speech-to-text software and the AI coding tool Cursor — a stack that lets operators mine stolen documents and draft phishing campaigns without ever shipping sensitive data to an outside AI service. Genians also found AI-generated finance and crypto decoy documents built to resemble legitimate investment reports, and says the group is moving beyond phishing lures toward malware development and attack automation. It is the offensive mirror of the defensive buildout we covered last week — Corma's $60 million raise for defensive cyber AI — Corma raises $60M for defensive cyber AI — and a reminder that cheap open-weight models have quietly erased the assumption that a sanctioned state cannot afford capable AI.
OpenAI is hiring a power-trading lead to run a commodity-hedging strategy across its data center electricity portfolio, Bloomberg reports — a role that reads like a trading desk, not a facilities department. The posting, which wants ten-plus years of power-trading or utility experience and deep knowledge of US power and natural gas markets, will "translate large, dynamic electricity and fuel exposures into practical hedging, procurement, and risk-management strategies," with a salary band of $181,000 to $285,000 plus equity. It is the financialization of the buildout in miniature: OpenAI is building a roughly $30 billion campus in Georgia, is reportedly in talks over a 10-gigawatt Ohio site, and paused a UK project over power prices — the same infrastructure push Wall Street is now underwriting at half-trillion scale, which we covered yesterday.
What to watch: whether OpenAI names a successor for Bakalar — and whether Kimsuky's offline AI stack shows up in a live campaign.
If labs keep losing their safety leaders, who decides what a model is allowed to do? Tell us in the comments.
