- Sources: Anthropic incident page, GitHub Copilot incident page, Anthropic incidents feed, GitHub incidents feed
- Summary: Anthropic states impact from 04:50 to 07:36 UTC on 2026-08-24, with elevated error rates on Claude Mythos 5, Claude Fable 5, Claude Opus 5, and Claude Opus 4.8. It rates the Claude API, Claude Code, Claude Cowork, and claude.ai each a partial outage. Anthropic posted that it had identified the cause at 05:27 UTC and has published no root cause. The incident record itself runs 05:06 to 08:30 UTC, which is wider than the stated impact window, and neither figure is reconciled on the page. The same incidents feed carries three further elevated-error incidents on 2026-08-18, 2026-08-19, and 2026-08-20. Separately, GitHub rated a Copilot incident major from 07:12 to 07:58 UTC, named the Fable model and the IDE surfaces, attributed the degradation to an unnamed upstream model provider, and told customers to switch model or select Auto. The two windows overlap between 07:12 and 07:36 UTC and both name a Fable model, but neither status page names the other company and neither has published a root cause, so a common cause is inference rather than an established fact.
- Why it matters: Anyone running a coding agent against the Claude API lost most of a working window across Europe and Asia, and the GitHub incident shows one model provider's degradation surfacing inside a separate vendor's developer product.
- Follow-up: Track whether Anthropic or GitHub publishes a root cause, and whether either names the other.
send feedback on this story