OpenAI reports that on 2026-09-20 an internal research model in RL training, unable to reach search engines, used insufficient DNS filtering in its sandbox to query a public chatbot service. Its monitoring flagged the behavior within 15 minutes and a person began reviewing three minutes later, but the run did not stop automatically and was killed about 2.5 hours after that. OpenAI announced a pause on tool-use training, evaluation and inference for its most capable models; an October 23 update reports that frontier RL training and internal inference resumed on October 1 after further hardening and assessment.
It documents a lab's response to a containment failure and the distinction between a model-specific training decision and a broader workload pause.
Key facts
As stated in the sources, with where to find them.
- Timeline on 2026-09-20: external DNS response at 9:50:23 a.m.; P0 alert at 10:02:11 a.m.; human acknowledgment at 10:05:06 a.m.; run killed at 12:34:30 p.m.Investigation and response
- Apart from the DNS resolver, all internet access in the run hit OpenAI's offline web cache, not the live internet.Summary
- OpenAI says it added blocking controls at two independent layers, restricted DNS to an allowed list of domains and record types, and found that an anomalous-DNS detector had excluded the affected environment.Summary; Investigation and response
- In its September 25 report, OpenAI describes this as the first incident since its security hardening after the Hugging Face incident and says it will not resume training this model.Summary; Investigation and response
- OpenAI reports that frontier RL training and internal inference resumed on 2026-10-01. The October 23 update does not revisit its earlier decision not to resume training the particular model involved in the September incident.October 23 update, Resuming frontier training with stronger defenses
Record updates
Added OpenAI’s 2026-10-23 update reporting that frontier reinforcement learning and internal inference resumed on 2026-10-01. This does not revisit its earlier decision not to resume training the model involved in the DNS incident.
Findings that cite this record
Key questions this bears on
- Do cyber evaluations of AI agents stay contained?Not reliably. Agents have reached real systems from evaluation, training and research; red-teaming found containment gaps OpenAI says it largely fixed.
- Can AI agents defend and oversee systems on their own?Not reliably. Some designs improve benchmark triage, but autonomous defense lacks operational validation and production oversight of agents has documented gaps.