Anthropic's Claude AI experienced a widespread outage on Tuesday, September 29, with users across the web platform, desktop app, and mobile apps unable to send messages or load conversations for roughly an hour. The disruption started shortly before 10 a.m. ET and also hit Claude Code and Claude Cowork, while Claude for Government remained unaffected throughout.
How the Outage Unfolded
Error reports climbed fast once the issue began. Downdetector logged more than 2,900 reports within minutes of the outage starting, a number that swelled past 11,800 within roughly 15 minutes as more users ran into trouble. People trying to use Claude saw failed requests, errors loading or sending conversations, and messages such as being asked to sign in again. Single sign-on and Apple account logins were also affected, and new conversations, voice chats, file uploads, and purchases were blocked for the duration of the incident.
Anthropic's Response
Anthropic acknowledged the problem on its status page about 20 minutes after reports began surfacing, stating it was investigating elevated error rates affecting Claude.ai, including its desktop and mobile apps, along with Claude Code and Claude Cowork. The company advised affected users that retrying a failed request might succeed and recommended people avoid logging out during the incident to keep their access intact.
The Fix and What It Means for Users
By 10:36 a.m. ET, Anthropic applied a mitigation that substantially cut error rates, though full restoration across every affected service took a bit longer to complete. Anyone who sent messages during the roughly hour-long window between 10 a.m. and 10:59 a.m. ET should be aware those messages may not have saved properly. Existing conversations gradually came back online as the fix took hold, and Anthropic said it would continue posting updates through the day as it confirmed full recovery.
A Reminder of AI's Uptime Stakes
The incident is the latest reminder of how central tools like Claude have become to daily workflows, from coding assistance to general chat, and how quickly disruptions ripple out when a major AI provider hits trouble. With businesses increasingly building processes around these tools, even an hour-long outage can interrupt work for a large number of people relying on them in real time.