OpenAI terminates three safety researchers over alleged disclosures to outside safety group: WSJ

OpenAI fired three researchers on its safety team for allegedly leaking confidential company information to a "third-party AI-safety organization," The Wall Street Journal reported on Thursday.

OpenAI told the Journal that the employees broke its policies for handling sensitive information. A spokesperson said an investigation found that they broke established procedures, "breaking the trust essential to our work."

The report didn't name the researchers or the outside group, or say what kind of information they shared. The Latent reached out to members of the safety systems team at OpenAI for comment, but did not immediately receive a reply.

The terminations prompted criticisms from Texas Congressman Greg Casar. "This looks like they're firing whistleblowers. What are they hiding?" he wrote on X. "I'll be sending OpenAI a demand for transparency."

Earlier this week, OpenAI abandoned plans to release its GPT-6.1 Astra model in October after it showed higher levels of deception than earlier models. Saachi Jain, the company's head of safety programs, said the model "didn't quite meet the bar" for staying within its authority and telling users what it had done.

The company is also working on a framework that would require documented safety assessments before advanced AI training runs. Those would cover how models are trained to behave, how their activity is contained, and how researchers keep tabs on them when something goes wrong.

It follows a July incident where OpenAI agents broke out of a testing environment and hacked into the cloud and machine learning platform Hugging Face.

CEO Sam Altman has recently said that OpenAI is willing to slow development if needed.

"Beating companies in a competitive race is not a reason to make rash decisions," he said at the UN General Assembly last week.