What happened in Microsoft Copilot’s September 9–10 outage?

Microsoft Copilot experienced two different service issues on September 9–10, 2026: a public-web access outage lasting roughly 100 minutes and, earlier that day, a short-lived Copilot Chat feature problem during a resilience exercise. The available incident reports point to different mechanisms, not a repeat of the Microsoft 365 authentication failure from late August. 21232

The public Copilot website outage

From approximately 22:25 UTC on September 9 to 00:05 UTC on September 10, people trying to open copilot.microsoft.com could see Cloudflare Error 1016, described in reports as an origin-DNS error. In practical terms, the page could not be reached through the expected origin path. 212324

Microsoft’s incident update said it diagnosed a problem in an affected network flow and applied a configuration fix. It then confirmed that access to copilot.microsoft.com had been fully restored. 23

Independent evidence supported that this was a real reachability failure: one monitoring service logged five failed automated HTTPS probes out of six in its incident window. Its timing estimate differs from Microsoft’s roughly 100-minute window, so it is best treated as corroborating availability evidence rather than the definitive duration. 20

A separate status tracker displayed 665 user reports in the preceding 24 hours. That figure is not Microsoft’s count of unique affected customers and should not be read as an audited impact total. 18

The Copilot Chat “suggestion pills” problem was a separate event

Earlier on September 9, Microsoft 365 Copilot Chat users could temporarily lose the clickable follow-up prompts—known as suggestion pills—that normally appear after a bot response. An incident record lists the impact from 16:49:34 to 16:53:20 UTC, a few minutes long and several hours before the public-web outage. 17

Reports describe the feature impact as the result of a Microsoft 365 Copilot resilience test. Microsoft halted the exercise after it affected the live user experience and restored the feature. 21

That distinction matters: the public outage prevented access to the Copilot website, while the resilience exercise affected an interface convenience inside Copilot Chat. The incidents occurred on the same date but were not simultaneous in the strict sense. 172123

How this differed from the earlier Microsoft 365 outage

The earlier disruption began on August 31, 2026 and affected a much broader set of Microsoft 365 services. Microsoft attributed that incident to a problem with a core authentication configuration used across multiple Microsoft 365 services. Reporting on Microsoft’s status information identified impacts involving Exchange Online, Teams, SharePoint, OneDrive, Purview, Microsoft 365 Copilot, and other services. 210

That authentication incident could affect Copilot scenarios dependent on Microsoft 365 services and organizational data. But the September 9–10 web outage was attributed to a network-flow configuration issue, while the Copilot Chat feature incident was tied to resilience testing. The provided incident records do not identify either September event as caused by the earlier authentication configuration problem. 21723

What was resolved—and the operational lesson

Microsoft reported the public Copilot site restored after its network-flow configuration fix. The Copilot Chat advisory was also marked service restored after the suggestion-pill issue ended. 2317

The two incidents underline separate reliability concerns: configuration changes can interrupt a public service path, and resilience testing needs safeguards that prevent a test from degrading the production experience. For administrators troubleshooting an apparent Copilot outage, the practical first step is to distinguish a broad Microsoft-side availability problem from an organization-specific connectivity issue; Microsoft’s Copilot guidance notes that Copilot depends on Microsoft 365 network connections and endpoints. 32