Claude outage: paying users lost access, while the government continued working — Anthropic's double standards
On June 23, 2026, an 85-minute outage of Anthropic's Claude AI paralyzed public access to the platform but left the government segment untouched. The incident revealed a fundamental division in infrastructure: while paid subscribers and API users lost money and time, government agencies continued to operate without a single interruption.
A signal of the problem arrived at 12:35 PM Moscow time. Anthropic specialists promptly launched an investigation and, after 35 minutes, stated they had found a technical solution. However, full restoration of all services took a full 85 minutes, and the "resolved" status appeared on the monitoring page only two hours after the first failure was recorded.
Five key products were hit: claude.ai, Claude API, Claude Code, Console, and Cowork. Notably, the specialized version of Claude for government agencies — Claude for Government — was not included in the official list of affected segments. This service, running on an isolated federal cloud platform from Palantir with FedRAMP High certification, continued to function normally.
The statistics speak for themselves
Data from the Anthropic status page over the last 90 days shows a glaring gap in service levels:
- Government version of Claude: uptime 99.93%, downtime approximately 90 minutes.
- Public service claude.ai: uptime 99.10%, downtime approximately 19 hours.
The difference in stability is nearly tenfold. And this is not a coincidence, but the result of a deliberate architecture. Last year, the General Services Administration (GSA) provided a similar level of access to all three branches of government for just $1. Fully isolated environments are considered the classic standard for strictly regulated cloud services. This is precisely why government AI systems continued to operate without interruption while consumer products faced large-scale outages.
The reaction from users, especially paid subscribers, was extremely harsh. One popular social media post accurately captured the audience's overall pessimistic mood: "Claude is down for everyone except the government."
The incident was just part of a very challenging period for the company's public segment. According to the official Anthropic status page, between June 9 and 23, technical specialists recorded more than 20 incidents. Users regularly encountered errors or complete system unavailability, which primarily affected the new flagship Opus 4.8.
To effectively cover the rapidly growing demand, the company promptly reserved up to 5 GW of new computing capacity from Amazon and signed new lease agreements for modern data centers. Against this backdrop, the isolated government segment initially remained safe, as it was not part of this overloaded resource pool.
My analysis: This architecture is not a mistake, but a conscious choice by Anthropic. However, for paid users who effectively subsidize the development and scaling of the infrastructure, such discrimination in service becomes a warning sign. If the company does not reconsider its priorities in resource allocation, trust in the commercial segment could be seriously undermined.