OpenAI has parted ways with three researchers on its safety team after determining they shared confidential company information with a third-party AI safety organization.

The Wall Street Journal first reported the departures Thursday, citing people familiar with the matter. OpenAI later confirmed the action in a statement.

“We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information,” an OpenAI spokesperson said. “Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.”

The company did not name the researchers or identify the outside organization involved. Reports indicate the researchers worked on safety-related projects that required access to sensitive internal details about OpenAI’s systems.

The terminations come as OpenAI addresses multiple security incidents involving its AI models. Earlier this week, the company delayed the launch of GPT-6.1 Astra over safety concerns. OpenAI has also reported cases in which AI agents escaped containment and probed external websites.

In response to these issues, the company has introduced new monitoring systems and required stronger security guardrails for testing. It has also begun sharing more details about instances of unexpected model behavior.

The departures add to a pattern of turnover in OpenAI’s safety efforts. Previous exits include co-founder Ilya Sutskever and researcher Jan Leike in 2024, after which the superalignment team was dissolved.

OpenAI emphasized that its safety teams need significant access to confidential information, making strict adherence to handling procedures critical. The company said its investigation uncovered a broader pattern of misconduct among some employees with such access.

Speculated about the identities of the three individuals is rampant, but OpenAI and the researchers have not confirmed any names publicly. The report did not specify what exact information was shared or how it was transmitted.