On Tuesday, Anthropic's Claude AI platform experienced a significant service disruption affecting both the web interface and API endpoints. According to Downdetector data, user error reports spiked past 2,000 by 1:27 p.m. ET before rapidly declining as Anthropic deployed a fix.

Anthropic's status page acknowledged the incident just before 1:30 p.m. ET, stating the team was investigating elevated error rates across multiple models, the Claude web interface, API, Claude Code, and Claude Cowork. Subsequent updates indicated a fix had been applied and monitoring was underway, though issues persisted with a subset of models.

From an engineering perspective, the rapid drop in user-reported errors suggests either a rollback or a targeted hotfix, but the lingering problems with specific models point to a more localized issue—possibly a model serving layer or inference optimization failure rather than a broad infrastructure collapse. The fact that both web and API paths were affected implies the root cause may sit in a shared dependency, such as authentication, routing, or model orchestration.

For teams relying on Claude for production workloads, this incident underscores the value of fallback strategies—caching common responses, maintaining alternative model providers, or implementing circuit breakers. The outage's short duration (roughly 30 minutes from peak to recovery) also highlights the importance of automated monitoring and rapid rollback pipelines.

This is a developing story. CNET has contacted Anthropic for further technical details on the root cause and remediation steps.

Source: CNET