On June 23, Anthropic's Claude experienced an 85-minute outage affecting public services including claude.ai, API, and Claude Code, while Claude for Government remained operational. The disruption began with elevated error rates across multiple consumer-facing models around early morning. Anthropic deployed a fix within 35 minutes, but normal service resumed after over two hours. The incident highlighted the deliberate infrastructure separation: Claude for Government runs on a FedRAMP High-authorized environment via Palantir's federal cloud, isolated from shared consumer systems. Over 90 days, the government tier logged 99.93% uptime versus 99.1% for claude.ai, a difference of roughly 19 hours of downtime. This gap fueled frustration among paying users already dealing with usage limits. Anthropic's status page recorded over 20 incidents between June 9-23, mainly involving its flagship Opus 4.8. To manage demand, Anthropic secured up to 5 gigawatts of new Amazon compute capacity and additional data center space, but the government tier never shared that strained pool.

Key facts

• 85-minute outage on June 23 affected claude.ai, API, Claude Code, Console, and Cowork.
• Claude for Government remained online due to FedRAMP High-authorized isolation.
• Over 90 days, government tier had 99.93% uptime vs 99.1% for consumer services.
• Anthropic logged over 20 incidents between June 9-23, mostly on Opus 4.8.
• Anthropic secured up to 5 GW of new Amazon compute capacity for rising demand.