Anthropic’s Claude suffered a global service outage on March 2, 2026, with problems centered on the claude.ai web app and login-related flows. Data from Anthropic’s status page and Downdetector showed thousands of user reports describing failed logins, prompt submission errors, and HTTP 500 responses.
Status updates pointed to frontend disruption, not the API
Anthropic marked the incident as Investigating at about 6:49 a.m. Eastern Time, or 11:49 UTC. At 12:06 UTC, the company said it was still looking into the issue. Then at 12:21 UTC, it posted the key update: Claude API operations were normal, while the visible failures were concentrated in the web experience and the login/logout path.
That left developers able to keep calling Claude models through API keys. The affected services included the claude.ai website, the platform.claude.com console, and frontend-related tools such as Claude Code. Anthropic classified the event as a partial outage. As of the latest update in the source material, the issue had not been resolved.
Thousands of reports surfaced across multiple regions
Downdetector showed a peak of thousands of incident reports. Users commonly described automatic logouts, inability to sign in, the message “Claude will return soon,” broken connections, and 500 internal server errors. In some cases, prompts could not be submitted at all.
Discussion spread quickly on Reddit and X. Reports came from North America, Europe, India, and Africa, suggesting the outage was not isolated to a single market or network segment but was global in scope.
Users relying on Claude had to pause or switch tools
People using Claude for work or study said the disruption had an immediate effect on their routines. Web access failures and console issues were enough to interrupt ongoing tasks. Some users moved to other AI services as a temporary workaround, while Anthropic continued investigating the cause without releasing a detailed explanation.