Claude suffers two hours-long outages in one day, Anthropic says

Anthropic’s Claude models went down twice in one day, with the first and larger outage lasting roughly seven hours, according to the company’s status page on August 5, 2026.

The first incident began around 07:05 UTC and hit Claude Opus 5, Claude Sonnet 5, Claude Mythos 5 and Claude Fable 5 with elevated error rates. Anthropic resolved it at 14:14 UTC after deploying a fix. A second, narrower disruption affecting only Claude Opus 5 started at 13:51 UTC and was cleared by 14:34 UTC.

Anthropic said it had identified the cause of the elevated errors but did not publish technical root-cause details. Users reported failed API requests, degraded performance in coding assistants built on Claude and problems reaching web apps, with some complaints starting as early as 3 a.m. ET.

The timing matters because Claude has become core infrastructure for a wave of AI coding tools, where a multi-hour API disruption stalls not just a chatbot but developers’ workflows and any product wired to Anthropic’s models. Outages of this length are a reliability test for a company selling itself to enterprises as a dependable backbone.

Anthropic did not say whether the two incidents shared a root cause. Whether it publishes a fuller post-incident review, and how often these disruptions recur, will shape how much enterprise buyers trust Claude for production workloads.