On July 29, 2026, Anthropic, the company behind the Claude AI assistant, posted a notice on its status page saying that all of its Claude models were experiencing "elevated errors." This meant the problem was not limited to one model or one type of customer. It affected the full lineup. Anthropic Status Page

This was not the first such incident in July. Earlier in the month, Anthropic recorded a separate problem affecting two specific models, Claude Opus 5 and Claude Haiku 4.5. That one was resolved by 14:34 UTC (7:34 AM Pacific). The scope was smaller than the July 29 event. Anthropic did not publish an explanation of what caused it. Anthropic Status History

There was also an outage on June 23, 2026, reported by TechRadar, which Anthropic confirmed was resolved. No official explanation for that disruption was provided either. TechRadar

Anthropic's status page, which tracks how reliably its service is running, recorded 99.40% uptime for claude.ai during July 2026. That sounds high, but for a service that businesses increasingly rely on, it translates to roughly 26 minutes of downtime or errors spread across the month. How much that actually affects people depends on when those errors happen, such as during a busy workday versus overnight. Anthropic Uptime

There is a meaningful difference between "elevated errors" and a full outage. A full outage means the service is completely unavailable. Elevated errors mean the service is still running but responding incorrectly, slowly, or with lower-quality results more often than usual. For teams that have built automatic backup systems (such as retrying failed requests or switching to a different provider when errors spike), the impact may be limited to slower performance. For teams that depend more directly on Claude without those safeguards, the errors can cause their own applications to fail, especially in systems where one step depends on the previous one completing successfully.

The jump from a two-model problem earlier in July to an all-model problem on July 29 is notable, though the available facts do not confirm whether the two events are related. Anthropic has not published an analysis connecting them. The June outage also lacked any official explanation.

The broader context here is that organizations relying on Claude should not treat the status page as their only source of information about reliability. The fact that all models were affected on July 29 suggests the problem may have been in shared infrastructure (the underlying systems that serve all models), meaning that switching from one Claude model to another for backup would not help if both rely on the same systems that are having trouble.

This kind of incident is not unique to Anthropic. Other major AI providers have experienced similar disruptions. The typical pattern goes like this: a problem is detected, a status page entry is posted, service is restored, and sometimes a follow-up explanation is published. Whether Anthropic publishes a detailed explanation for the July 29 event will be informative. The lack of explanation for the June outage left customers without the information they would need to improve their own backup strategies.

The 99.40% July uptime figure is not disastrous, but it falls below the 99.9% or higher reliability target that many business contracts expect. Teams that have promised their own customers a certain level of availability, based on assumptions about how reliable Claude would be, may find that figure uncomfortable. The practical lesson is that having a backup plan involving a different provider entirely, not just a different model from the same provider, continues to be worth the effort and cost.

For now, the July 29 incident page is the most reliable record of what happened. Whether Anthropic follows up with a detailed explanation or closes the matter quietly, as it did with the June event, will determine how much customers can learn from it.