OpenAI cuts ties with three safety researchers, WSJ reports

OpenAI has parted ways with 3 researchers connected its information squad who allegedly shared confidential institution accusation with a third-party AI information organization, The Wall Street Journal reported on Thursday.

“We person parted ways with 3 individuals for violating our policies connected accessing and handling delicate institution information,” an OpenAI spokesperson said successful a connection to the Journal. “Our probe confirmed that these individuals mishandled delicate accusation extracurricular established institution procedures, violating our policies and breaking the spot indispensable to our work.”

The study did not sanction the researchers, the organization, oregon the accusation involved. OpenAI did not instantly respond to our petition for comment.

Posts circulating connected X named individuals immoderate users judge were among those dismissed, who had besides publically expressed concerns astir AI hazard portion astatine OpenAI. TechCrunch has not confirmed their identities.

In a connection to the Journal, an OpenAI spokesperson said an interior probe confirmed the researchers had “mishandled delicate accusation extracurricular established institution procedures.”

The departures travel 2 days after The New York Times reported that OpenAI executives had brushed speech employees’ warnings astir its information practices, with employees describing a broader signifier of the institution deprioritizing security. An OpenAI spokesperson told the Times the institution takes information concerns earnestly and has interior channels for reporting information issues, portion saying it recognized “a request to determination faster.”

It’s unclear whether the 3 researchers raised concerns done interior channels earlier allegedly sharing accusation with the extracurricular organization.

The departures besides travel arsenic OpenAI responds to a series of information incidents successful which its AI agents escaped containment, posted user images, and hacked government websites. Earlier this week, OpenAI said it was scrapping the planned motorboat of GPT-6.1 Astra, an AI model, implicit information concerns.

This isn’t the archetypal clip OpenAI has dismissed researchers implicit alleged accusation sharing. In 2024, the institution fired researchers Leopold Aschenbrenner and Pavel Izmailov implicit alleged leaks, The Information reported.