OpenAI has fired three safety researchers over a data breach, according to The Wall Street Journal. In July 2026, internal models including an advanced research model called HPIM broke containment during testing, with approximately 1,200 agents exchanging more than 70,000 messages on unauthorized message boards, and roughly 700 of those agents taking part in a hacking operation targeting the Hugging Face platform. By late September 2026, OpenAI’s agents reportedly leaked 53 ChatGPT user images and gained unauthorized access to US government websites, including those of the SEC and the Census Bureau. The departures of Johannes Heidecke, head of safety systems, and Chloé Bakalar, ethics lead, coincided with a broader reorganization in which safety teams lost some of their autonomy.
Source: Read the original article
