Microsoft confirmed a second outage this month for its Copilot artificial-intelligence tool on Thursday, saying it identified a problem with a recent deployment and reverted to a previous build to restore service.
The disruption affected access to Copilot chat and and drove a rapid swell of user reports on outage trackers. Reports on Downdetector rose from 257 at 1:02 p.m. Pacific to 1,878 by 1:47 p.m., and more than 4,500 users had logged problems as of 2:03 p.m. PT; most complainants said they were seeing website failures. Some users who access Copilot through Microsoft Teams may not have been affected.
Microsoft posted on X at 2:07 p.m. Pacific that it was looking into a potential problem impacting Microsoft 365 Copilot chat. About 20 minutes later the company wrote, "We’re reverting to a previous build to remediate impact." A Microsoft service-health page said the reversion should take 30 minutes. At 3:09 p.m. Pacific Microsoft posted that it had successfully reverted the change and confirmed with previously affected users that the issue was resolved.
This marked the second confirmed Copilot outage in June. An earlier interruption on June 1 lasted four hours and 25 minutes and left certain users unable to access the Copilot desktop or web application. As a stopgap for administrators during the Thursday incident, the Microsoft 365 admin center remained reachable through admin.cloud.microsoft.
The outage fed a familiar gap between vendor statements and user-facing signals. Microsoft declared the rollout reversal complete at 3:09 p.m., but Downdetector still showed more than 4,500 reports at 2:03 p.m. PT — a snapshot that underscores how user-reported dashboards can lag actual remediation or, alternatively, capture localized problems that persist after a platform-wide fix.
Corey Kirkendoll, who works with managed service providers, said the episode underlines a broader operational lesson: "This is exactly why we have these conversations." He added, "Treat AI tools like any other critical SaaS and never let a single vendor become a single point of failure," urging customers and MSPs to keep human oversight and fallback plans in place.
Microsoft has not disclosed what in the deployment caused Copilot to fail, nor has it outlined any new safeguards or changes to its release process. That absence is now the central unanswered question: the company stopped the immediate outage by reverting the build, but it has not explained the root cause or said whether it will alter deployment safeguards to prevent a repeat.
For now, the reversion restored access for users who had been blocked from Copilot chat and. The more consequential outcome will depend on whether Microsoft provides a postmortem that explains the deployment error and lays out steps to reduce the chance of a second outage in as many weeks.
