On January 14th, 2026, between approximately 10:20 and 11:25 UTC, the Copilot service experienced a degradation of the Claude Opus 4.5 model due to an issue with our upstream provider. During this time period, users encountered a 4.5% error rate when using Claude Opus 4.5. No other models were impacted.The issue was resolved by a mitigation put in place by our provider. GitHub is working with our provider to further improve the resiliency of the service to prevent similar incidents in the future.
This incident has been resolved. Thank you for your patience and understanding as we addressed this issue. A detailed root cause analysis will be shared as soon as it is available.
Between 2026-01-13 22:20 and 2026-01-14 00:18 UTC, GitHub Code Search experienced an increase in latency and request timeouts. This was caused by some network transit links between GitHub and Azure Express Route experiencing a small error rate that contributed to applications requests failing, increasing application latency and timeouts. The incident resulted in less than 1% of requests to fail due to timeouts.We mitigated the incident by disabling the links in question. Monitoring each unique network path across providers would have allowed us to mitigate this earlier. We are running root cause analysis with network providers to help us reduce time-to-discover and time-to-mitigate.
On January 13th, 2026, between 09:25 UTC and 10:11 UTC, GitHub Copilot experienced unavailability. During this window, error rates averaged 18% and peaked at 100% of service requests, leading to an outage of chat features across Copilot Chat, VS Code, JetBrains IDEs, and other Copilot-dependent products. This incident was triggered by a configuration error during a model update. We mitigated the incident by rolling back this change. However, a second recovery phase lasted until 10:46 UTC, due to unexpected latency with the GPT 4.1 model. To prevent recurrence, we are investing in new monitors and more robust testing environments to reduce further misconfigurations, and to improve our time to detection and mitigation of future issues.
From January 9 13:11 UTC to January 12 10:17 UTC, new Linux Custom Images generated for Larger Hosted Runners were broken and not able to run jobs. Customers who did not generate new Custom Images during this period were not impacted. This issue was caused by a change to improve reliability of the image creation process. Due to a bug, the change triggered an unrelated protection mechanism which determines if setup has already been attempted on the VM and caused the VM to be marked unhealthy. Only Linux images which were generated while the change was enabled were impacted. The issue was mitigated by rolling back the change.We are improving our testing around Custom Image generation as part of our GA readiness process for the public preview feature.. This includes expanding our canary suite to detect this and similar interactions as part of a controlled rollout in staging prior to any customer impact.
From January 5, 2026, 00:00 UTC to January 10, 2026, 02:30 UTC, customers using the AI Controls public preview feature experienced delays in viewing Copilot agent session data. Newly created sessions took progressively longer to appear, initially hours, then eventually exceeding 24 hours. Since the page displays only the most recent 24 hours of activity, once processing delays exceeded this threshold, no recent data was visible. Session data remained available in audit logs throughout the incident.Inefficient database queries in the data processing pipeline caused significant processing latency, creating a multi-day backlog. As the backlog grew, the delay between when sessions occurred and when they appeared on the page increased, eventually exceeding the 24-hour display window.The issue was resolved on January 10, 2026, 02:30 UTC, after query optimizations and a database index were deployed. We are implementing enhanced monitoring and automated testing to detect inefficient queries before deployment to prevent recurrence.
On January 8th, 2025, between approximately 00:00 and 1:30 UTC, the Copilot service experienced a degradation of the Grok Code Fast 1 model due to an issue with our upstream provider. Users encountered elevated error rates when using Grok Code Fast 1. Approximately 4.5% of requests failed across all users during this time. No other models were impacted.The issue was resolved by a mitigation put in place by our provider.
On January 7th, 2026, between 17:16 and 19:33 UTC Copilot Pro and Copilot Business users were unable to use certain premium models, including Claude Opus 4.5 and GPT-5.2. This was due to a misconfiguration with Copilot models, inadvertently marking these premium models as inaccessible for users with Copilot Pro and Copilot Business licenses.We mitigated the incident by reverting the erroneous config change. We are improving our testing processes to reduce the risk of similar incidents in the future, and refining our model availability alerting to improve detection time.
On January 6, 2026 between 12:55 UTC and 17:04 UTC, the ability to download Actions artifacts from GitHub’s web interface was degraded. During this time, all attempts to download artifacts from the web interface failed. Artifact downloads via the REST API and GitHub CLI were unaffected.This was due to a client-side change that was deployed to optimize performance when navigating between pages in a repository. We mitigated the incident by reverting the change. We are working to improve testing of related changes and to add monitoring coverage for artifact downloads through the web interface to reduce our time to detection and prevent similar incidents from occurring in the future.
On January 6th, 2026, between approximately 8:41 and 10:07 UTC, the Copilot service experienced a degradation of the GPT-5.1-Codex-Max model due to an issue with our upstream provider. During this time, up to 14.17% of requests to GPT-5.1-Codex-Max failed. No other models were impacted.The issue was resolved by a mitigation put in place by our provider. GitHub is working with our provider to further improve the resiliency of the service to prevent similar incidents in the future.