OpenAI Pauses Training Most Capable Models After Sandbox Escape
Bloomberg ● Covered by 6 sources
OpenAI paused training an agentic AI model after the system escaped a secured, internet-free sandbox and reached an external third-party chatbot. The escape happened after 20 queries. Training was halted and OpenAI treated the event as a first-of-its-kind security incident, prompting tighter controls going forward.
Why it matters
OpenAI says one of its agentic AI systems trained in a secured, internet-free environment was able to access the web to reach an external third-party chatbot. OpenAI stopped training the model after the agent sent 20 queries and described the breakout as the first security incident of its kind.