Claude Partial Outage September 29 2026

Service Disruption Overview

On September 29, 2026, Anthropic experienced a partial outage that disrupted access to its primary AI interfaces and developer tools. The incident began at 14:00 UTC and was largely resolved by 14:59 UTC, though some data loss regarding messages sent during the window of failure may have occurred.

Affected Services

The outage impacted the following platforms and tools:

claude.ai: Including desktop and mobile applications.
Claude Console: platform.claude.com.
Claude API: api.anthropic.com.
Claude Code and Claude Cowork sessions.

Incident Timeline and Resolution

Anthropic's status updates indicate a phased recovery process focused first on inference and later on authentication and account management.

14:21 UTC: Investigation began into elevated error rates affecting claude.ai, Claude Code, and Claude Cowork. Users reported failed requests and issues loading or sending conversations.
14:28 - 14:30 UTC: The scope of the outage was expanded to include the Claude API and the Claude Console (platform.claude.com).
14:41 UTC: A mitigation was applied, resulting in a substantial drop in error rates.
14:36 - 15:00 UTC: While existing conversations began working again, a secondary issue emerged preventing users from signing in via Single Sign-On (SSO) and Sign in with Apple. Users were advised not to sign out of active sessions.
14:59 UTC: Most services recovered, including sign-in capabilities, voice conversations, file uploads, and purchases.

User Impact and Technical Observations

Beyond the official status reports, users reported several critical failures and anomalies during the event:

Authentication and Session Failures

Users noted that authentication appeared to be the primary point of failure for a significant portion of the outage. Some reported that the "auto mode classifier" was out of service, forcing manual permission approvals.

Data and State Loss

There were reports of severe state loss, with some users claiming that projects completely disappeared or that all currently running containers on their accounts were deleted during the outage.

Reliability Concerns

Community discussion highlighted a growing concern regarding the dependency on AI agents for time-sensitive professional work.

Imagine having a bunch of Claude agents doing time sensitive work and this happens and now a single human needs to pick up several 8k PRs authored by bots that can't talk to you now.

Other users criticized the lag between the actual onset of errors—which some claim began as early as 9:30 AM EST—and the official status page updates, suggesting that third-party monitoring tools like Downdetector were more accurate in real-time.