OpenAI fires three safety researchers over suspected leaks

OpenAI has fired three safety researchers, according to the Wall Street Journal. The company says they mishandled sensitive information. OpenAI suspected they leaked internal model designs to Ryan Greenblatt, an AI safety researcher, in part through disappearing Signal chats.

What was reported

Based on the report, OpenAI believes the three researchers shared details of how its AI models are built. Model designs describe the inner workings of an AI system. Companies treat them as closely held secrets.

Some of the sharing allegedly happened on Signal, a private messaging app. Disappearing chats delete messages after a set time, which can leave little record behind.

Why it matters

AI companies hold powerful technology and guard it hard. Safety researchers, meanwhile, often want the public to know more about how these systems work and what risks they carry. This case sits right where those two goals collide.

For everyday users and business owners, the question is how much outsiders can know about the AI tools they rely on. Tight secrecy protects a company's work. It can also limit how much others can check.

What is not known yet

• What exactly was shared, and how much of it.
• Whether the three researchers dispute OpenAI's account.
• Whether Ryan Greenblatt has responded.
• Whether any further action will follow.

The material available so far comes from a short summary of the Journal's reporting. Suspicion is not proof, and the details may change as more comes out.