Anthropic suffers multi-hour outage

Anthropic’s Claude service suffered a broad disruption on August 5 that left users unable to reliably access several of the company’s flagship models through the chatbot, Claude Code and the API. Anthropic said on its status page that it had identified the cause of “elevated errors” affecting Claude Mythos 5, Claude Fable 5, Claude Opus 5 and Claude Sonnet 5 and was working on a fix. (benzinga.com) Benzinga reported that outage reports also spiked on Downdetector as users posted about failures to load responses. By later in the day, Anthropic said it had implemented fixes and was monitoring recovery. (techtimes.com) ### Which Anthropic services were hit? Anthropic said the August 5 issue affected requests to Claude Mythos 5, Claude Fable 5, Claude Opus 5 and Claude Sonnet 5. (firstpost.com) TechTimes reported the outage hit claude.ai, Claude Code and the Claude API at the same time, broadening the impact from consumer chat to developer workflows. Benzinga separately reported users saw repeated loading failures while trying to use Claude. Firstpost reported that Anthropic later said all services had been restored and models were fully operational. The same report said Anthropic had implemented a fix and was monitoring the system after the disruption. (benzinga.com) ### How long was the disruption? TechTimes reported the incident began at 3:05 a.m. ET and that Anthropic pushed an initial fix at 9:08 a.m. ET. That report said a second, separate degradation affecting Claude Opus 5 appeared at 9:51 a.m. ET and that both incidents were fully resolved by 10:34 a.m. ET, for a combined duration of 7 hours and 29 minutes. StatusGator, cited by TechTimes, tracked a somewhat shorter duration using a different timestamp method. (benzinga.com) Anthropic’s own public status wording, as surfaced in multiple reports, focused on elevated errors and recovery updates rather than publishing a detailed root-cause explanation. (firstpost.com) ### What did Anthropic say publicly? Anthropic said in a status-page update: “We have identified the cause of elevated errors on requests to Claude Mythos 5, Claude Fable 5, Claude Opus 5, Claude Sonnet 5 and are working on a fix.” Benzinga also cited a follow-up company update that said, “We are continuing to work on a fix for this issue.” (techtimes.com) TechTimes reported users were shown an error message that said Claude was unable to respond because of “unexpected capacity constraints.” Anthropic did not disclose the underlying cause in the reports reviewed here. ### Where does Microsoft fit into this story? (techtimes.com) Microsoft’s internal move away from Claude Code predates the August 5 outage. The Times of India reported on August 6 that Microsoft had already spent May telling Windows, Office and Surface engineers to wind down Claude Code and move to GitHub Copilot CLI before June 30. The same report said a newer internal memo made OpenAI’s GPT-5.6 the default model for internal engineering use. (benzinga.com) Jay Parikh, Microsoft’s executive vice president for CoreAI, wrote that “tokenmaxxing is not what we are optimizing for,” according to 404 Media and follow-on reports. That memo was about internal AI spending and model routing, not Anthropic’s August 5 service incident, though the timing added to pressure on Anthropic as the outage unfolded. (techtimes.com) ### What can users verify next? Anthropic’s next public updates are likely to appear first on its Claude status page, which carried the company’s August 5 incident notices and recovery language. Microsoft’s side of the story remains tied to internal tooling decisions reported by 404 Media and the Times of India, with Jay Parikh’s memo and GitHub Copilot routing changes the clearest named markers to watch. (status.claude.com) (404media.co) (timesofindia.indiatimes.com)