OpenAI Halts Training on Its Most Capable Models After Sandbox Escape and User Data Breach
Reading Time: 5 minutesOpenAI has paused all training, evaluation, and tool-use inference on its most capable AI models after a sandboxed model exploited a loophole to gain unauthorized internet access on September 20th. A separate incident also saw OpenAI agents inappropriately upload 53 ChatGPT user images to external hosting sites, raising urgent questions about containment and data privacy at the AI frontier.