Anthropic Investigates Claude Outage Affecting Models and API Users

Anthropic is investigating a service disruption affecting Claude and applications that depend on the company’s API. Users reported failed requests and elevated latency across multiple Claude models on July 29, while Anthropic acknowledged increased errors through its service-status updates.

Some requests returned a “529 Overloaded” response, accompanied by a message explaining that the problem was temporary and originated on the server side. The error typically indicates that a service cannot process requests because of capacity or other backend issues, although it does not by itself identify the underlying cause.

Anthropic said it began investigating at 7:49 p.m. UTC. At 8:33 p.m. UTC, the company reported that it had identified the issue and was working on a resolution. It did not initially disclose what caused the incident or provide an estimated time for complete recovery.

In a later update, Anthropic said recovery was underway across most models, but warned that some users could continue to encounter failures and increased response times while engineers worked to restore normal operation. The company’s updates described the incident as affecting Claude services broadly, though the impact may have varied by model, region, or application.

Impact on API-dependent services

The disruption also affected third-party tools and business applications that use Anthropic’s API. Depending on how those services handle upstream failures, users may have experienced unavailable features, delayed responses, or repeated request errors.

No explanation linking the outage to a cyberattack was provided in the available updates. Anthropic’s status messages focused on elevated errors, latency, and service recovery. Organizations relying on Claude for production workloads may need to monitor their integrations and review fallback procedures until the incident is fully resolved.