Fired OpenAI Safety Researchers Deny Leak Claims
Three safety researchers recently fired by OpenAI have publicly disputed the company’s allegations that they leaked confidential information, according to reporting by The Wall Street Journal cited by The Decoder. Tomek Korbak, Mikita Balesni, and Jasmine Wang stated they were never informed in writing about which specific data they supposedly mishandled.
The firings followed an internal investigation into communications using encrypted messaging apps and discussions regarding AI model transparency. Investigators examined chats with external researchers concerning the intermediate reasoning steps of advanced artificial intelligence models, a central topic in debates over safe oversight. Wang additionally faced scrutiny over an accidental email access incident she reported herself.
The researchers told the Journal that the terminations and subsequent investigations create a chilling effect across the industry. They warn that such actions may discourage independent safety auditors and internal staff from collaborating openly or raising critical risk concerns with external organizations. In a letter directed to OpenAI leadership, the trio has urged the company to establish closer cooperation with external safety groups to ensure proper oversight of frontier systems.
Based on reporting by the-decoder.com.
