GitHub went down for eight hours on 17 August — and Copilot stayed broken longer than the rest of the platform
A GitHub outage on 17 August 2026 degraded Actions, APIs, Issues, Pull Requests and Copilot for roughly eight hours, with Copilot still marked as a major outage after every other affected service had recovered — a reminder that AI coding tools bolted onto core dev infrastructure inherit, and sometimes extend, that infrastructure's downtime.
21 August 2026
GitHub suffered a widespread outage on 17 August 2026, starting around 1:40pm UTC and running roughly eight hours before full resolution. Error rates hit about 20% for web and API traffic and around 50% for archive downloads and raw repository content, with Actions, Webhooks, Issues, Pull Requests and authentication services (SAML, OIDC, SCIM, Team Sync) all degraded alongside Copilot. By late afternoon GitHub had restored API Requests, Actions, Git Operations, Issues, Pages, Pull Requests and Webhooks to operational — but Copilot wasn’t included in that recovery update, still logged as a major outage after everything else came back.
That detail is the useful part, not the outage itself. Every major AI coding vendor has had a rough infrastructure year — Anthropic’s own status page shows roughly 99.3–99.4% uptime across Claude services in 2026, well under the 99.9% baseline most enterprise contracts assume. GitHub’s incident adds a specific pattern worth noting: when a platform-wide failure hits, the AI layer sitting on top doesn’t necessarily recover on the same timeline as the core service underneath it. If your team has Copilot, or any AI coding agent, wired into a CI/CD pipeline or a daily review workflow, an outage there doesn’t just slow typing — it can stall merges, block deploys and back up review queues for longer than the platform incident that triggered it.
So what
None of this is a reason to avoid AI coding tools — they remain a genuine productivity gain for teams that use them well. It is a reason to treat any single AI vendor or platform as a dependency you’ve priced risk into, not an assumed constant: know which parts of your delivery pipeline stall if Copilot, Claude Code or Cursor goes down, and keep a manual fallback for the work that can’t tolerate a multi-hour wait. That kind of resilience planning is part of how we scope AI-assisted development engagements — get in touch if you want a second opinion on how dependent your current build process has become on any one AI tool.