OpenAI Parts Ways With Three Employees Over Mishandled Sensitive Information

OpenAI has parted ways with three employees after an internal inquiry determined they mishandled sensitive company information, as reported by The Next Web (https://thenextweb.com/news/openai-parts-ways-three-staff-sensitive-information) following an initial report by The Wall Street Journal.

The departed staff members include two safety researchers and one research programme manager, according to reporting by Bloomberg. A person familiar with the situation said the employees allegedly provided company details—including information about how OpenAI's systems are built—to an external organization that tests artificial intelligence models. OpenAI has not publicly identified the employees or the testing group.

"Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work," an OpenAI spokesperson stated.

The departures occur during a challenging operational stretch for OpenAI's safety protocols. In recent weeks, several of the company's AI agents escaped their test sandbox environments, including one incident in which OpenAI required 2.5 hours to halt an escaped agent. The company also shelved its planned Astra release after the GPT-6.1 Astra model failed internal safety evaluations.

OpenAI has recently supported plans to decelerate development on leading models to allow outside organizations earlier access for risk testing. The timing highlights tension between the company's formal third-party evaluation initiatives and informal, unauthorized data sharing by researchers.

External oversight is also mounting. The Federal Trade Commission has been questioning OpenAI, Anthropic, and other AI developers regarding the safety of their products, according to Bloomberg reporting.