Anthropic Service Outage Reports

Anthropic’s official status page shows an “Investigating” update at Mar 26, 23:06 UTC and an “Identified” update at Mar 27, 00:49 UTC specifically naming elevated errors on Opus 4.6 and Sonnet 4.6. (anthropic.statuspage.io) Business Today described the March 27 incident as the fifth Claude disruption in March, citing user-report percentages for Chat and Code impact, while an industry tracker counted as many as eight disruptions for the month and reported non‑government services falling below 99% uptime. (businesstoday.in) (aiproductivity.ai) Crowd-sourced monitors recorded large user-report volumes during the episode, with IsDown logging roughly 846 reports in a 24‑hour window and StatusGator recording about 552–559 user‑submitted reports around the same period. (isdown.app) (statusgator.com) Anthropic’s incident history lists operational side‑effects during outages—delays to billing, purchasing API credits, and access shutoff/restore workflows—which the company later marked as resolved. (status.claude.com) Reporting also noted the web UI and authentication paths were hit hardest while some API and enterprise integrations continued to function for customers. (ibtimes.com.au) Anthropic posted time‑stamped updates on its status site throughout the incident but, according to industry commentators, has not published a comprehensive post‑mortem beyond those incident updates as of late March. (anthropic.statuspage.io) (aiproductivity.ai) Technical writeups and engineering posts documenting the disruptions are pushing concrete resilience patterns for production systems now: multi‑provider LLM fallback chains, automated provider switching and retry logic, and provider‑agnostic routing libraries to reduce single‑vendor failure blast radius. (markaicode.com) (deployflow.co)