Claude and Anthropic’s API Go Down for Nearly an Hour in September 29 Outage
Anthropic’s Claude platform experienced a partial outage on September 29, 2026, that began at 14:00 UTC (7:00 a.m. PT) and knocked out sign-ins, new conversations, voice features, and file uploads across claude.ai, the Claude API, Claude Code, and Claude Cowork. The company’s status page shows the incident was substantially mitigated by 14:36 UTC, about 36 minutes in, with error rates dropping sharply, and full recovery confirmed at 14:59 UTC. Anthropic has not disclosed a root cause.
What Happened
Table of Contents
According to Anthropic’s status page, the incident manifested as “elevated error rates” across multiple services simultaneously rather than a single feature failing. Users trying to sign in were blocked, existing sessions that tried to start new conversations failed, voice input and output stopped working, and file uploads errored out. The Claude API — used by third-party developers building on Anthropic’s models — was affected as well, meaning the disruption extended beyond the consumer chat app into production applications built on top of Claude.
Anthropic’s own update noted that “some messages sent between 14:00 and 14:59 UTC may not have been saved,” meaning a subset of users may have lost in-progress conversation data even after service was restored. The company did not publish a technical post-mortem identifying whether the cause was a code deployment, a dependency failure, a capacity issue, or something else — a pattern consistent with how Anthropic has handled previous incidents this year.
Why It Matters
Claude has moved well beyond a consumer chatbot. Claude Code is used by software teams for day-to-day development work, and the Claude API sits underneath a growing number of commercial products, from customer support agents to the kind of agentic shopping and enterprise-safety integrations covered elsewhere on this site. When Anthropic’s infrastructure goes down, the outage doesn’t just interrupt a chat session — it can halt automated coding workflows, break live customer-facing agents, and stall business processes that companies have built assuming near-continuous uptime.
A 40-to-60-minute window is short by the standards of major cloud outages, but for teams running production agents or CI/CD pipelines that call the Claude API, even a brief spike in error rates can cascade into failed jobs, retried requests, and downstream costs. The fact that sign-ins and new-chat creation were both affected also suggests the failure sat somewhere central in Anthropic’s authentication or request-routing layer, rather than in a single downstream feature like voice.
Timeline of the Incident
- 14:00 UTC — Anthropic detects and reports elevated error rates across Claude services.
- 14:00–14:36 UTC — Sign-ins, new chat creation, voice features, and file uploads fail intermittently for affected users; the Claude API also returns errors.
- 14:36 UTC — Error rates drop substantially as mitigation takes effect, roughly 36 minutes after onset.
- 14:59 UTC — Anthropic confirms full recovery, just under an hour after the incident began.
Anthropic’s status updates described services as “operating normally since recovery,” with no follow-up degradation reported in the hours after.
A Pattern of Outages in 2026
This is not an isolated event. Anthropic experienced a separate outage on August 16, 2026, that began at 21:58 UTC and ran for roughly 42 minutes, affecting claude.ai, Claude Code, Claude Cowork, and platform.claude.com with login failures and degraded performance, while the Claude Console and Claude API remained unaffected that time. In both cases, Anthropic acknowledged the disruption publicly and restored service within the hour, but declined to publish a root-cause explanation.
The recurrence of short, sharp outages across different combinations of services — sometimes hitting the API, sometimes sparing it — points to reliability challenges that come with scaling an AI platform whose usage has expanded rapidly across consumer, developer, and enterprise customers simultaneously. As Claude is increasingly wired into third-party commerce, coding, and safety tooling from companies like Shopify and Nvidia, the operational bar for uptime continues to rise.
Frequently Asked Questions
How long did the Claude outage on September 29 last?
The outage began at 14:00 UTC. Anthropic’s status page shows error rates dropped substantially by 14:36 UTC, about 36 minutes later, with full recovery confirmed at 14:59 UTC — just under an hour after it started.
What caused the Anthropic outage?
Anthropic has not disclosed a root cause. Its status page described the issue only as “elevated error rates” affecting authentication, new conversations, voice features, file uploads, and the Claude API, without further technical detail.
Were any messages or chats lost?
Anthropic warned that some messages sent between 14:00 and 14:59 UTC may not have been saved, meaning a subset of users could have lost in-progress conversation content even after the service came back online.
Related Articles
- Anthropic Says Its Claude Agents Discovered a New Enzyme System in Phage DNA
- Amazon Opens Its Seller Tools to Anthropic’s Claude in New AI Agent Push
Helpful Resources
- Anthropic Status Page — live incident and uptime history for Claude services
- Anthropic — official company site
