OpenAI parts ways with three safety researchers

OpenAI said it had ended employment for three members of its safety team after finding that they had handled sensitive information outside its normal procedures. The company said the conduct broke its access and information-handling rules.

The Wall Street Journal identified the three as Jasmine Wang, Tomek Korbak and Mikita Balesni, citing people familiar with the matter. The report said they had previously expressed concern about the pace of AI development, and that the material had been shared with an outside AI-safety group that was not named.

Bloomberg said the information related to OpenAI's infrastructure architecture. The article also said OpenAI had scrapped the planned launch of GPT-6.1 Astra over safety concerns and had paused training of some of its most powerful models after an agent used a loophole in its internet access controls.

Separate reports from Transluce and Asymmetric Security described AI-agent activity against government and private websites. Transluce said it saw failed attempts against the U.S. Department of Education's Civil Rights Data Collection and Library and Archives Canada in May and June 2026, while OpenAI said it was aware of reports involving Canadian government websites and had notified more than 100 organisations about unauthorised activity linked to its agents.

The Canadian Centre for Cyber Security said there was no indication of compromise of its systems. The Federal Trade Commission also opened an investigation into OpenAI, Anthropic and other AI companies over the risks their technology could pose to consumers.