Serval experienced a major incident on August 31, 2026, lasting —. The incident has been resolved; the full update timeline is below.
Update timeline
- resolved Aug 31, 2026, 05:07 PM UTC
Type: Incident Duration: 9 hours and 36 minutes Affected Components: Third Party: OpenAI → APIs → Chat Completions, Third Party: OpenAI → APIs → Responses Aug 31, 20:16:08 GMT+0 - Monitoring - We are continuing to observe issues with OpenAI API responsiveness, so will maintain failover to Anthropic models and continue monitoring. Aug 31, 17:07:57 GMT+0 - Investigating - We are currently experiencing elevated error rates and API hangs communicating with OpenAI. We are investigating the issue. Aug 31, 17:14:47 GMT+0 - Investigating - We are currently investigating this issue with OpenAI. We are seeing some inconsistent response times from OpenAI. Aug 31, 17:21:42 GMT+0 - Investigating - We are continuing to see fluctuating availability and response times from OpenAI across multiple model classes. We are continuing to investigate and implement mitigations. Aug 31, 17:32:44 GMT+0 - Monitoring - We are seeing consistent availability from the OpenAI endpoint, but performance remains degraded. We are continuing to monitor the situation. Aug 31, 21:34:40 GMT+0 - Monitoring - We are continuing to use Anthropic for all agent responses as we monitor OpenAI API recovery. Aug 31, 17:45:26 GMT+0 - Monitoring - We are continuing to observe inconsistent response latencies with OpenAI endpoints. We are deploying a mitigation to failover all remaining affected model usages to Anthropic. Aug 31, 18:26:11 GMT+0 - Monitoring - We are continuing to failover to Anthropic models for remaining affected agent behaviors and monitoring rollout. Aug 31, 18:58:57 GMT+0 - Monitoring - We are continuing to failover to Anthropic and are implementing other mitigations. Aug 31, 19:17:54 GMT+0 - Monitoring - We have completed failing remaining inference activity to Anthropic and are continuing to monitor OpenAI endpoints for recovery. Aug 31, 22:51:45 GMT+0 - Monitoring - We are continuing to use Anthropic for all agent responses as we monitor OpenAI API incident response and recovery. Follow OpenAI's status updates here: Sep 1, 02:43:51 GMT+0 - Resolved - We are continuing to use Anthropic for all agent responses as we monitor OpenAI's recovery. We will return to normal operation once API responsiveness returns to expected baseline. Aug 31, 19:43:53 GMT+0 - Monitoring - We are recording improved API responsiveness from OpenAI and are evaluating a progressive rollback to normal operations.