OpenAI fires 3 safety researchers over alleged leaks, WSJ reports
OpenAI confirmed violations of its rules on sensitive information, the Wall Street Journal reports, after firing alignment researchers Jasmine Wang, Mikita Balesni and Tomek Korbak. David Robinson also left.
The Wall Street Journal reported on 2 October that OpenAI fired three safety researchers, Jasmine Wang, Mikita Balesni and Tomek Korbak, according to The Decoder's account. They allegedly leaked confidential information to an outside AI safety organisation. OpenAI confirmed "violations of the company's rules for handling sensitive company information", but did not confirm the names of those fired.
A fourth researcher, David Robinson, has also left the safety team, and The Decoder describes his exit as a fourth departure. Nothing in that account establishes a link to the firings, and it does not say why he went. This paper has not seen the Journal's original report.
Much remains unclear. The report as relayed does not say what information was shared or which organisation received it. The Decoder notes that Korbak had worked with METR and Redwood Research on security testing of AI agents, but says no link between those groups and the alleged leaks has been confirmed.
The Decoder also lists earlier public remarks by the four, which it says appeared in September. Korbak voiced unhappiness with OpenAI's direction, Balesni put the odds of AI causing human extinction above 10 per cent, and Wang signed a petition for slower AI development. Robinson agreed that "the race toward self-improving AI might be insane". The account does not say these statements played a part in the firings.