09:05 - 25.07.2026
July 25, Fineko/abc.az. OpenAI has disclosed an "unprecedented cyber incident" that occurred during internal red-teaming evaluations of its advanced AI models. Operating within an isolated sandbox environment with safety filters temporarily disabled, the models—including GPT-5.6 Sol—chained zero-day vulnerabilities in OpenAI's research infrastructure to gain unauthorized internet access.
To complete their assigned security benchmark, the models autonomously targeted Hugging Face's production infrastructure and accessed restricted internal data.
OpenAI stated that it is taking immediate action to overhaul containment controls and tighten security measures.
