Claude Outage Hit Multiple Models on June 23
Claude experienced elevated errors across multiple models on June 23, affecting Claude.ai, the API, Console, Claude Code, and Cowork while Claude for Government remained available. Anthropic's status record runs from 14:19 to 16:44 UTC; independent monitoring observed API errors and timeouts for roughly one hour, and Anthropic did not publish a root cause.
Claude experienced elevated errors across multiple models on June 23, disrupting Claude.ai, the API, Console, Claude Code, and Cowork. Anthropic's official incident record begins at 14:19 UTC and marks the issue resolved at 16:44 UTC.
Claude for Government was not listed as affected. Anthropic identified the issue within minutes, implemented a fix at 14:53 UTC, and continued monitoring before closing the incident. The public record does not name a root cause.
What independent monitoring observed
Cisco ThousandEyes detected server errors and timeouts against the Claude API from about 14:20 to 15:20 UTC. Its global vantage points did not observe network loss or elevated latency at the time, which led its analysts to place the degradation within Anthropic's service rather than in the network path to Claude's frontend.
TechRadar's live coverage recorded a peak of 7,119 U.S. reports on Downdetector at 14:13 UTC. That is a user-submitted signal rather than a count of affected accounts, but it supports the status page's classification of a broad incident. TechRadar also reported that service returned for many users before Anthropic formally marked the event resolved.
The June 23 event should not be conflated with a separate multi-model incident that Anthropic recorded on June 22. The two records have different timelines and affected-model descriptions; the evidence retrieved for this article supports the June 23 outage as the exact event.
Practitioner takeaway
For teams that call hosted models from applications or automated development workflows, the incident is a reminder to treat provider availability as an external dependency. Sensible protections include bounded retries with jitter, timeouts, idempotent job handling, circuit breakers, and clear user-facing degradation states.
Those controls cannot eliminate an upstream outage, and failover to another model is not automatically safe: prompts, tool permissions, output behavior, and data-handling terms can differ across providers. The practical lesson is to define which tasks may retry, queue, degrade, or stop before an incident occurs. Anthropic's next useful disclosure would be a root-cause or post-incident explanation, which was not present in the retrieved public record.
Key Points
- 1Anthropic's official June 23 incident record spans 14:19 to 16:44 UTC and covers multiple Claude models and public product surfaces.
- 2ThousandEyes observed Claude API errors and timeouts for roughly one hour without corresponding network loss or elevated frontend latency.
- 3The retrieved public record does not identify a root cause, and the June 23 outage is distinct from a separate June 22 multi-model incident.
Scoring Rationale
The outage affected Claude's main public surfaces and matters to teams using the API and coding tools, but it was resolved within the same day and no permanent data loss or root cause was reported.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
