Anthropic outage on October 23, 2026
Measured window 23:41 UTC to 00:13 UTC on October 23, 2026, 32 minutes. Worst status observed: major.
StatusBird recorded Anthropic as non-operational from 23:41 UTC on 23 October 2026. The worst reading was major. Checks run every two minutes, and 16 readings in a row were non-operational. Service looked recovered at 00:13 UTC, a measured duration of 32 minutes. Anthropic posted an incident at 23:37 UTC titled "Elevated errors for Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5".
Incident at a glance
| Service | Anthropic |
|---|---|
| Date | October 23, 2026 |
| StatusBird first saw it | 23:41 UTC |
| Vendor posted an incident | 23:37 UTC |
| Recovery observed | 00:13 UTC |
| Duration | 32 minutes |
| Worst status observed | major |
| Non-operational readings | 16 |
All times UTC. StatusBird polls Anthropic every 2 minutes from its own infrastructure, so the true start and end sit within 2 minutes of the times above.
Was Anthropic down on October 23, 2026?
Yes. StatusBird recorded Anthropic as major from 23:41 UTC until 00:13 UTC, a measured 32 minutes. 16 consecutive checks came back non-operational.
What caused the Anthropic outage?
The cause is not known. Anthropic has published no root cause statement that StatusBird can read. The vendor's status page described the problem as elevated errors for three named models, but that is a description of the symptom, not an explanation. If Anthropic publishes a cause later, it will not change the readings above, which are what StatusBird measured from outside.
Anthropic published an incident titled "Elevated errors for Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5" at 23:37 UTC. A root cause has not yet been published by the vendor, so nothing on this page attributes one.
How long was Anthropic down?
32 minutes, measured from the first non-operational reading at 23:41 UTC to the first reading that cleared it at 00:13 UTC. StatusBird counts a service down from the first reading it fails, not from the moment the vendor admits it, which is why this number can be larger than the one on the vendor's own incident report.
Timeline, in UTC
| Time (UTC) | Event |
|---|---|
| 23:37 UTC | The vendor posts an incident on its own status page |
| 23:41 UTC | StatusBird records its first non-operational reading |
| 23:44 UTC | A second consecutive non-operational reading confirms the outage, alerts go out and this page is published |
| 00:13 UTC | StatusBird records the first reading that clears the outage |
Who and what was affected
This hit any part of your store that sends requests to Claude Fable 5.1, Sonnet 5.5 or Haiku 5.5 during that 32 minute window. In practice that means things like AI chat replies to shoppers, automated first-line support responses, generated product descriptions and tags, and review or ticket summarising. If one of those steps sits inside a flow a customer is waiting on, the customer saw an error or a blank result rather than a slow one.
How StatusBird detected it
StatusBird recorded its first non-operational reading for Anthropic at 23:41 UTC, and Anthropic's own status page confirms the incident, posted at 23:37 UTC. StatusBird polls every 2 minutes from its own infrastructure, so these are two independent records of the same window.
Method: StatusBird reads Anthropic's published status endpoint every 2 minutes and records every reading. Availability figures count major and critical only. The full method is at statusbird.io/methodology.
Lessons for merchants
- Put AI calls behind a timeout and a fallback, so a failed call shows a static message or hands the shopper to a person instead of breaking the page.
- Keep AI out of the checkout and payment path, so an outage costs you content and replies rather than orders.
- Queue background AI work, such as product copy and tagging, and retry it later rather than dropping the job.
- Log your own error rates per vendor and alert on them, since your requests can start failing before any status page is updated.
The longer version, with the checks specific to this service, is in the Anthropic outage playbook.
Frequently asked questions
Was Anthropic down on October 23, 2026?
Yes. StatusBird recorded Anthropic as major from 23:41 UTC on October 23, 2026, for a measured 32 minutes.
How long was Anthropic down?
32 minutes, measured from the first non-operational reading at 23:41 UTC to the first clear reading at 00:13 UTC. Readings are taken every 2 minutes, so the true start and end sit within 2 minutes of those times.
What caused the Anthropic outage?
Anthropic published an incident titled "Elevated errors for Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5" at 23:37 UTC. A root cause has not yet been published by the vendor, so nothing on this page attributes one.
How was this Anthropic outage detected?
StatusBird recorded its first non-operational reading for Anthropic at 23:41 UTC, and Anthropic's own status page confirms the incident, posted at 23:37 UTC. StatusBird polls every 2 minutes from its own infrastructure, so these are two independent records of the same window.
Was my store affected?
If your store calls Anthropic during checkout or fulfillment, the requests to check are the ones made between 23:41 UTC and 00:13 UTC on October 23, 2026. Orders created in that window without a matching downstream record are the usual casualty.
Monitor Anthropic with StatusBird
This page exists because StatusBird was already watching. It checks Anthropic and the rest of the stack an online store runs on every 2 minutes and sends an SMS, email, Slack, Teams or Discord alert the moment a reading goes non-operational, which on this incident was 23:41 UTC.
Start monitoring free Live Anthropic status All Anthropic outages