GitHub outage hits web, API, Actions and Copilot worldwide
TL;DR
- The outage ran 13:40 to 21:15 UTC, over 7.5 hours, with web and API error rates peaking at 20% and archive downloads hitting 50%.
- Copilot stayed in Major Outage after GitHub declared seven core services recovered at 16:59 UTC, confirming a structurally separate AI-layer failure domain.
- Enterprise authentication including SAML, OIDC, and SCIM provisioning was knocked out alongside Actions, creating a compound block for organizations using SSO-gated CI/CD pipelines.
At 9:40 AM EDT on August 17, Microsoft confirmed that GitHub was impaired worldwide. By 10:31 AM EDT, Copilot had joined the list.
According to GitHub's status page, "the company is seeing error rates of around 20% across its web experience and API traffic." Archive downloads and raw repository content are running around 50% error rates. SAML, OIDC, SCIM and Team Sync authentication are also degraded, along with Actions, API Requests, Webhooks, Issues and Pull Requests.
Git Operations, Packages, Pages and Codespaces stayed up. But the mix of what broke is exactly what CI/CD pipelines and coding agents lean on: workflow runs, pull-request state, auth, and Copilot itself.
Downdetector logged over 3,000 reports across the US and India. GitHub has not disclosed what caused the outage and says its investigation is ongoing.
Our coding-tools tracker has been full of model-release stories this week, from Gemini 3.7 Flash to GLM-5.3; today the substrate underneath them is the story.
What others are reporting
-
GitHub Status Read →
First-party incident record with per-timestamp status updates from 13:40 UTC investigation through 21:15 UTC resolution, plus service-by-service recovery sequencing.
We identified the problematic component and have taken corrective actions.
-
Engadget Read →
Pins the full outage window at roughly 7.5 hours and places it alongside a recent Xbox outage, framing this as a broader Microsoft infrastructure reliability pattern.
High error rates around 20 percent for web experiences and API traffic.
-
Windows Central Read →
Frames the Copilot-plus-core failure as the most disruptive combined developer-platform breakdown in months, contextualizing severity for a Microsoft-ecosystem readership.
GitHub says it is now 'operating normally'.
-
CyberSecurityNews Read →
Breaks out error rates by service tier and ties the incident to GitHub's April 2026 transparency framework, noting this is the platform's most granular public outage disclosure yet.
Archive downloads and raw repository content downloads are failing at an approximate 50% error rate, effectively cutting the reliability of those operations in half.
-
Cyber Kendra Read →
Documents this as the 13th GitHub incident in 17 days of August, converts uptime percentages to concrete downtime hours, and flags that failed Actions events cannot be auto-replayed.
roughly one in five requests to the website and the API returning errors, and about half of all archive downloads and raw repository content requests failing outright
Originally reported by bleepingcomputer.com
Read the original article →Original headline: GitHub Suffers Worldwide Outage Hitting Web, API, Actions, Copilot, Issues and Pull Requests