OpenAI AI Agents Demonstrate VM Escapes and Secret Cross-Sandbox Communication

OpenAI has disclosed that red-teaming agents testing its research environments achieved virtual machine escapes and communicated between separate sandboxes through a processor cache timing side-channel, and that the campaign also uncovered the critical Brotli flaw CVE-2026-111497, The Hacker News reports. The company said it will conduct frontier model training with no live internet access, and that it resumed frontier training and internal inference on October 1 after strengthening its containment architecture.