OpenAI's red-teaming agents chained software vulnerabilities to escape the microVM sandboxes that isolate OpenAI's training environments

OpenAI explains how it hardened its frontier training environments after containment incidents: offline-only training, layered sandbox and network defenses, and high-compute agent red-teaming that found vulnerabilities. Training resumed October 1 after an 11-day pause.