A configuration error in the service that verifies users of Microsoft 365 entered its third day Wednesday with no full recovery date, even as email service largely came back. Microsoft said most mail is flowing again but left parts of its collaboration suite, its file services and its AI assistant running in a degraded state.
The disruption began Aug. 31 at 3:08 p.m. UTC. Microsoft traced the cause to a misconfiguration in a core authentication system shared by many of its cloud services, according to company status updates.
The service at the center of the failure is the identity layer that sits in front of Microsoft’s cloud, the system that decides who can enter any product with a single set of credentials. Now branded Microsoft Entra ID, it replaced the directory service that once guarded only corporate networks; today it guards mail, documents, meetings and AI assistants at once. When it is healthy, users sign in once and reach everything. When it is not, every product that depends on it feels the effects.
The most essential traffic recovered first. Mail flow through Exchange Online was restored to more than 99% of normal levels, Microsoft said, and most customers could send and receive messages again as the company worked through the affected configuration.
The rest of the suite moved more slowly. Search remains degraded, Teams channels and calendar functions are impaired, OneDrive and SharePoint file synchronization is still catching up, and the Copilot features that ground answers in a user’s mailbox data are running in reduced form, the company said.
Microsoft said it has entered an extended monitoring period and will publish its next status update at 4:30 p.m. UTC on Sept. 2. It gave no estimate for when the remaining services will be fully restored, saying only that it would keep customers informed as the affected systems stabilize.
The breadth of the failure says something about how Microsoft 365 is built. Mail, documents, meetings and the AI assistant all check the same identity and configuration layer before letting anyone in, so a single faulty setting can ripple across products that look unrelated to each other.
The dependence is broad. Microsoft 365 counts its paid commercial seats in the hundreds of millions, and for many companies the suite is effectively the office itself, holding the calendar, the documents, the chat history and the mail archive. An outage that leaves search slow and file sync stalled touches work that was supposed to have moved beyond such failures years ago.
The episode tests enterprises that have pushed their daily work deep into the suite. Companies that run their communications and files on Microsoft’s cloud have few workarounds for the affected services, and the outage leaves IT teams tracking status pages while employees work around slow search and stalled file synchronization.
Copilot’s degradation carries its own message. The assistant’s answers depend on data from mail and files, so when the underlying systems wobble, the AI wobbles with them. When the mailbox is healthy enough to read but not fully synchronized, the grounding that makes Copilot useful is the first thing to suffer.
Microsoft has weathered broad cloud outages before, and each one renews questions about concentrating so much work in one vendor’s stack. The company has pushed customers toward a single suite covering mail, meetings, files and AI, betting that integration outweighs the risk of shared failure. This week’s outage is the downside of that bet.
Microsoft’s identity service has broken before. A January 2021 outage in the same family of directory services locked users out of email and meetings for hours and became one of the most scrutinized incidents in the company’s cloud history. This week’s failure is narrower in scope, but it touches more products at once, including the AI assistant Microsoft has made the centerpiece of its productivity pitch.
Corporate customers may also have claims on Microsoft. Enterprise agreements typically include service credits when availability falls short of contracted levels, though the accounting for a partial, days-long degradation is rarely quick or simple.
Outages rarely move market share by themselves, analysts said, but they accumulate in the accounts that procurement teams keep. A disruption that touches mail, meetings, files and an AI assistant at once gives enterprises more to remember than a single service failure would.
The Copilot angle gives the outage a second audience. Microsoft has sold 365 Copilot to enterprises as an assistant that knows their mail and files, and this week those same customers watched its answers degrade because the systems it reads were unreliable. The episode is a practical lesson in how AI features inherit the fragility of the data beneath them, and enterprises piloting AI on top of Microsoft’s stack are taking notes.
For now the company’s task is technical: restore the remaining services, watch the authentication layer through the extended monitoring period, and keep the recovery from introducing new problems as traffic returns to normal levels.
For customers, the practical question is how long the degraded services take to return. Microsoft has said it will keep updating the status page; until the monitoring period ends, enterprises that rebuilt their daily workflows around 365 have little to do but wait.
