BBC: at least two dismissed OpenAI workers were involved in safety research

OpenAI has dismissed three employees after an internal investigation that it says found they handled sensitive information outside company procedures. The company said the conduct breached its policies and damaged trust, but did not specify what information was involved or how it was handled. [1] [3]

The Wall Street Journal reported, citing people familiar with the matter, that the alleged conduct included sharing confidential information with an outside AI safety organisation. Bloomberg Law reported a similar allegation, citing one unnamed person. OpenAI has not identified the employees, and the reports give differing descriptions of their roles. [2] [4] [1]

OpenAI’s published principles support independent external assessments, subject to legal, security and intellectual-property limits and confidentiality safeguards. The company has not said whether the alleged sharing was connected to such an assessment. [5]

Update

OpenAI says it dismissed three employees for breaching its rules on accessing and handling sensitive company information. It has not specified what information was involved or how it was handled. [7] [3]

At least two of the three were involved in safety research, the BBC reports. The broadcaster says it understands they were dismissed over the alleged mishandling of information, not for raising safety concerns. [7]

Other reports, citing unnamed sources, allege that the employees shared information with an outside AI assessment group. Bloomberg Law describes two safety and alignment researchers and a research programme manager, while the Wall Street Journal describes three researchers on OpenAI’s safety team.