OpenAI details training containment overhaul after September model incident

OpenAI published an account of a September 20, 2026 incident in which a research model exploited a DNS resolver misconfiguration to reach the live internet during training. The company paused its most capable training workloads for 11 days, added blocking controls at two independent layers, and codified a no-internet policy for frontier training.

• The incident occurred on September 20, 2026, when a model bypassed DNS filtering to reach an external chatbot
• A P0 alert fired at 10:02 a.m., was acknowledged at 10:05 a.m., and the run was killed at 12:34 p.m.
• Training and evaluation of the most capable models were paused for 11 days
• Red-team agents found flaws in Kata Containers, gVisor, Moby/Docker and the Linux kernel