Claude partial outage

Access to Claude's services, including its API, is currently unreliable for many users.

On September 29, 2026, Anthropic experienced a significant partial outage affecting various services. The incident began around 14:00 UTC with elevated error rates across claude.ai, Claude Code, Claude Cowork, and the Claude API. By 14:36 UTC, initial errors were largely mitigated, with existing conversations mostly restored. However, a secondary issue emerged, preventing many users from signing in, starting new chats, or accessing features like Claude Code and Cowork, alongside an inability to make purchases or upload files. The company stated that users who were already signed in should avoid logging out. Updates indicated that while error rates dropped substantially after a mitigation was applied, some sign-in attempts and account actions continued to fail, with a full fix still in progress.

The incident highlights the fragility of AI service availability and its direct impact on user workflows, especially for those reliant on continuous access for tasks like coding or collaborative work. The dual nature of the outage, first with general errors and then with login and feature-specific issues, suggests a complex underlying problem that may require more than a simple fix. This disruption could lead users and businesses to re-evaluate their reliance on a single AI provider and explore more robust multi-cloud or multi-model strategies for critical operations.

With AI models becoming increasingly integrated into daily workflows, such outages underscore the need for better resilience and fallback options. For Anthropic, this event is a reputational challenge and a call to action for enhanced infrastructure stability and more transparent communication during incidents. Users should anticipate a period of increased vigilance regarding service uptime from all major AI providers, potentially influencing future purchasing decisions and architecture designs to accommodate such risks.