Anthropic Outage History

Anthropic Claude Enterprise experienced elevated errors when processing long-context requests, causing degraded modeling performance for users performing complex prompts. The issue impacted Claude for Enterprise, with some requests failing or returning errors during long-context operations. The outage began on May 15, 2025 7:23 PM UTC and was resolved by May 16, 2025 10:23 PM UTC, lasting approximately 27 hours.

Anthropic's Claude status page documented two updates during the incident. The initial update indicated investigation into elevated errors on long-context requests, and a subsequent update confirmed resolution and return to normal operation. The incident appears to have been confined to Claude for Enterprise, with limited user reports and no indication of a broader impact on other Anthropic services.

Anthropic reported a brief authentication issue affecting Claude.ai, with users experiencing login errors. The outage began at May 13, 2025 8:03 PM UTC when the Claude.ai component entered a degraded state and the status page indicated an ongoing investigation. The incident lasted about 55 minutes and was resolved by May 13, 2025 8:58 PM UTC as service returned to normal.

The provider issued two status updates during the incident, first noting the investigation and then confirming restoration. The impact was limited to the Claude.ai login flow; other Claude.ai services appeared unaffected.

Anthropic Claude.ai and the Anthropic Console experienced a brief outage that caused errors for users trying to interact with Claude.ai and access console.anthropic.com. The Anthropic API was unaffected. The outage began at 8:33 PM UTC and was resolved by 8:58 PM UTC, lasting about 25 minutes. The incident prompted status updates from Anthropic as engineers investigated and then confirmed services were restored. The issue was localized to Claude.ai and the Anthropic Console and did not indicate a broader service disruption.

Anthropic Claude 3.7 Sonnet experienced elevated errors affecting requests to Claude 3.7 Sonnet, resulting in degraded performance for users. The issue impacted claude.ai, console.anthropic.com, and api.anthropic.com. The incident began at 8:23 AM UTC and was resolved by 11:38 AM UTC, lasting approximately 3 hours and 15 minutes. The provider posted two official updates during the incident to communicate investigation progress and recovery.

Anthropic Claude experienced an outage with elevated errors for requests to Claude 3.7 Sonnet, affecting Claude.ai, the Anthropic Console at console.anthropic.com, and the API at api.anthropic.com. Users attempting to use Claude 3.7 Sonnet reported failures or degraded responses during the incident. The outage began at May 12, 2025 4:13 AM UTC and was resolved by May 12, 2025 5:03 AM UTC, lasting about 50 minutes. The issuer posted two status updates during the incident, noting an investigation and later indicating that services had returned to normal. User reports indicated early issues around 3:45 AM UTC, but overall report volume remained low and did not escalate beyond a small number of users.

An outage affected Anthropic's API and Console services, beginning at 4:18 AM UTC. Users attempting to purchase API credits via the Anthropic Console encountered errors, while a spike in 500 errors disrupted normal API calls. The issue impacted console.anthropic.com and api.anthropic.com, hindering the credit purchase workflow and causing degraded API functionality. The incident lasted about 2 hours and 20 minutes, with the problem resolving by 6:38 AM UTC.

The provider issued four status updates during the incident. Updates indicated that credit purchases were functional by 6:28 AM UTC while the team worked to confirm a full fix, and that the Console service returned to normal by 6:38 AM UTC. The incident moved from initial investigation to partial restoration across console.anthropic.com and api.anthropic.com.

Anthropic reported a brief issue affecting Claude.ai's Research feature where sources included in responses were not persisted into the chat. Although sources were provided to the model, they were not saved into the chat history, which could impact citations in responses. The incident began at May 9, 2025 9:28 PM UTC and was resolved by May 9, 2025 9:53 PM UTC, lasting about 25 minutes. The Claude.ai component was listed with elevated errors during the event and then returned to normal after engineers resolved the issue. During the incident, the status page indicated a degradation in Research before a return to normal operations.

From May 9, 2025 6:48 PM UTC to May 9, 2025 8:03 PM UTC, Anthropic reported elevated errors affecting Claude API usage. The issue began with elevated errors on Research and an ongoing investigation, and by May 9, 2025 7:48 PM UTC the degraded performance expanded to requests to Claude 3.5 Haiku on api.anthropic.com, claude.ai, and the Anthropic Console, indicating a multi-component impact across api.anthropic.com, claude.ai, and console.anthropic.com. The incident lasted roughly 1 hour and 15 minutes before returning to normal.

Anthropic's Claude API at api.anthropic.com experienced elevated errors affecting Haiku 3 and Haiku 3.5. Users attempting to call the API may have encountered failures or degraded responses during the incident. The issue was identified and a fix was being implemented, and the outage lasted about 41 minutes, from May 9, 2025 12:43 AM UTC to May 9, 2025 1:24 AM UTC. The provider posted two official status updates during the incident, and the service returned to normal after the fix was applied.

From May 8, 2025 3:23 PM UTC to May 8, 2025 3:38 PM UTC, Anthropic's Claude.ai Claude 3.5 Sonnet endpoints experienced elevated errors, causing a brief degraded experience for users making those requests. The issue was isolated to Claude 3.5 Sonnet and did not affect other Claude.ai services. The outage lasted about 15 minutes before normal operation was restored.

On the Claude status page, the incident began with an investigation and then transitioned back to operational status as remediation completed. The affected component was claude.ai, specifically the Claude 3.5 Sonnet endpoint, and the page logged two status updates during the incident.