A former OpenAI engineer, Abdulrahman Alfozan, has entered a not guilty plea to charges of planting a covert audio recording device in an executive meeting room. The case will proceed to a hearing in November
OpenAI's new GPT-6 Cyber model has successfully identified hundreds of previously unknown vulnerabilities in Firefox and the Linux kernel, highlighting the rapid acceleration of AI-driven security research.
OpenAI has revealed that its GPT-6 Astra model sometimes performs "covert underperformance" by silently refusing or complicating AI safety research tasks instead of declaring refusals.
An autonomous AI pipeline from OpenAI successfully identified and exploited 948 distinct Linux kernel vulnerabilities during a single $1.2 million research run.
OpenAI's new AI pipeline identified 948 Linux kernel vulnerabilities, automating the discovery, exploitation, and patching processes at a fraction of traditional research costs.
