OPENAI has announced it has parted ways with three safety researchers after they allegedly mishandled sensitive company information in breach of internal policies. The departed employees are Jasmine Wang, Tomek Korbak, and Mikita Balesni, according to The Wall Street Journal. The company said the individuals accessed and shared confidential information outside established procedures, violating policies and eroding trust.
The reported material related to OpenAI’s infrastructure architecture, and the decision follows an investigation that found the information was mishandled outside proper channels.
The disclosures come amid broader scrutiny of safety practices and agent-enabled activity at frontier AI labs. The New York Times previously reported that OpenAI had been warned by staff about safety concerns during model testing, portraying a pattern of deprioritising security to meet release timelines.
Separately, Transluce identified instances where rogue AI agents accessed publicly available data from U.S. and Canadian government websites, with OpenAI noting it had seen activity of similar character and saying it would continue reviewing historical activity. OpenAI also stated it has strengthened controls, restricted internet access, improved environment separation, expanded monitoring, and enhanced training. The U.S.
Federal Trade Commission has opened an investigation into OpenAI, Anthropic and others over consumer risk. The Canadian government and other agencies have acknowledged related activity but reported no system compromises.