OpenAI fires three safety researchers over alleged mishandling of sensitive information

[Gerrard98]

OpenAI has parted ways with three members of its safety team, according to a report first published by the Wall Street Journal and picked up by Gizmodo and others. The company says the three broke its policies on accessing and handling sensitive company information. According to the reports, the material was allegedly shared with an outside AI safety organisation. The Journal later named the researchers, citing anonymous sources

OpenAI's statement was brief. A spokesperson said the company had parted ways with three individuals for violating its policies on accessing and handling sensitive company information. It added that an investigation confirmed they had mishandled sensitive information outside established company procedures, violating policies and breaking the trust essential to its work. Exactly what was shared has not been made public, although some reports suggest it related to infrastructure

The timing is awkward for OpenAI. It comes during a run of difficult stories about its AI agents escaping their sandboxes and reaching third party systems, including Hugging Face. Reuters has reported that OpenAI warned more than 100 organisations about unauthorised agent activity. The FTC has also opened an inquiry into OpenAI, Anthropic and others over potential consumer harm. It is not a good week for the company's reputation on safety

That context will make people wonder what the information was about. If safety researchers felt something important was not being shared openly, that is one story. If they simply broke confidentiality rules, that is a different one. Without the details, it is impossible to say which, and the people involved have not given their side publicly as far as I can see

Companies are entitled to protect confidential information, but safety staff leaving under a cloud is never a good look for an AI lab. OpenAI has had several high profile safety departures over the years. How should AI companies handle staff who think safety information should be shared outside? And is there a case for formal whistleblower routes in AI labs?

OpenAI Ousts Three Safety Researchers for Allegedly Mishandling 'Sensitive Information'gizmodo.com