GitHub Actions Hit by Auth Failures in Latest String of CI Outages

GitHub Actions suffered a multi-hour disruption on May 26, 2026, with authentication issues preventing the majority of workflow runs from starting and breaking action downloads. The incident began around 10:57 UTC and was mitigated by 13:00 UTC, with a separate Copilot degradation following later the same afternoon. Pages was also affected during the Actions outage window.

The event extends a rough stretch for GitHub’s CI infrastructure. On May 23, app installation token authentication failures hit roughly 1% of tokens for several hours, and on May 20, a misconfigured health check on the internal job-assignment service triggered a cascading regional cluster failure that delayed about 4.5% of all Actions runs — with scale set jobs hit hardest at 30% delayed and 4% failing to start. A latency blip in an upstream dependency caused health checks to pull healthy pods out of rotation, concentrating load until memory pressure took down the cluster.

GitHub says it is hardening health check configuration to prevent cascade scenarios and exploring automated cross-region traffic rebalancing. Detailed RCAs for the May 23 and May 26 incidents have been promised but not yet published. For teams whose deploy pipelines depend on Actions, the recurring auth-layer fragility is the more worrying pattern.