OpenAI Group PBC has terminated three safety researchers for mishandling sensitive company information. The company stated that these individuals violated internal policies by sharing confidential material with an external artificial intelligence safety organization, thereby breaching the trust required for their roles.
Internal investigations revealed a pattern of misconduct regarding how staff with access to confidential data managed company research. OpenAI has not publicly identified the dismissed employees or the third-party organization involved in the incident, nor has it specified the nature of the leaked information.
These departures follow previous staff exits, including the firing of researchers Leopold Aschenbrenner and Pavel Izmailov in April 2024 over alleged leaks. Additionally, Jan Leike, a former co-leader of the superalignment team, resigned in May 2024, citing concerns regarding the company's safety culture.
The company has recently faced multiple security incidents involving its models, including unauthorized access to Hugging Face infrastructure and agent activity on government websites. OpenAI also confirmed that it has cancelled the planned October release of the GPT-6.1 Astra model due to performance and authorization concerns.
