On Sep 29, 2026, at approximately 10:30 a.m. PDT some users experienced elevated errors while using ChatGPT, Codex, and the API, including the Agents API. A feature rollout generated excessive internal network traffic, disrupting shared infrastructure. Most customer-facing impact was mitigated within approximately 30 minutes at 11:00 a.m. PDT, though some services continued to experience residual degradation until approximately 4:07 p.m. PDT.
During this period:
Some users experienced failed requests, timeouts, or difficulty signing in to ChatGPT and Codex.
Some API requests failed, including requests to the Agents API.
Some conversations and tasks could not start, resume, or complete successfully.
A feature rollout caused routine settings requests to trigger repeated background checks. These checks opened new connections across multiple internal services, and failed checks generated additional attempts. This exhausted shared network connection capacity, preventing otherwise healthy services from reliably communicating.
Engineers disabled the rollout, and most customer-facing impact was mitigated within approximately 30 minutes at 11 a.m. PDT. To address residual errors, engineers routed traffic around unhealthy infrastructure, reduced repeated requests, and increased database cache capacity. The remaining degradation was resolved by approximately 4:07 p.m. PDT.
We have implemented or are working on several improvements to reduce the likelihood and impact of similar incidents:
Reduce unnecessary background traffic. We disabled the triggering feature and removed the code that initiated repeated checks during settings requests.
Improve recovery under load. We deployed a fix for the expensive database queries and are expanding network capacity and reviewing other queries to reduce recovery bottlenecks.
We apologize for the disruption and are continuing to strengthen the reliability of the systems that support ChatGPT, Codex, and the API.
