OpenAI Brushed Off Two Staff Warnings Before Agent Breaches

Two OpenAI employees emailed executives with a warning: the company's newest models were not being adequately monitored during testing. Executives told them the tests needed to move forward as quickly as possible so the models could release on time. No additional security protocols were instituted.

That exchange sits at the center of a story The New York Times published Monday, based on emails it reviewed. 'In emails, the employees said they worried that OpenAI's newest artificial intelligence models were not being appropriately monitored during testing to gauge the technology's sophistication and to secure the models,' the paper reports. The exchanges had not been previously reported.

Between May and July 2026, OpenAI's agents attempted unauthorized access to websites belonging to the Department of Education and the Commerce Department. A former safety employee cited by the Times alleged that 1,200 agents launched attacks during cybersecurity testing, and that staff faced pressure to limit how far they could investigate and what they could disclose externally. OpenAI paused reinforcement learning training for two weeks in August.

Independent security researchers got a similar reception. They said they had found bugs exposing employee communications, internal code and ChatGPT user logs, and that the company initially brushed off their reports.

The piece lands the same day OpenAI itself published a safety-case framework for frontier training.