A disruption of Microsoft services such as Outlook, Teams and Copilot stretched into Asia-Pacific business ‌hours on Friday after affecting thousands of US users, as the tech giant struggled to fix server problems it blamed for the outage, Reuters reported.

The outage hit government agencies in Australia, Japan and the United States, as well as companies including OpenAI and Anthropic. It is the second disruption to Microsoft’s 365 services this year, raising concerns that repeated failures could push corporate clients to question Microsoft’s reliability.

In its latest update, Microsoft said it had applied adjustments to the affected servers, but they did not result in sustained recovery. It said it had observed some partial recovery in the regions where it applied the adjustments.

“We’re reviewing what additional actions we can take to improve overall infrastructure stability,” the company said on X.

Downdetector, which tracks outages by collating status reports from a number ⁠of sources, showed more than 18,000 reports of Outlook problems in the US at 11:30 p.m. ET (0430 GMT). It also showed outages for several Microsoft services in the Asia-Pacific region, including India, Australia and Japan, early Friday local time.

At its peak, Downdetector showed more than 130,000 reports of Outlook problems in the US around 1 p.m. ET on Thursday. The outage has lasted for about 12 hours.

In a service health notice, Microsoft said features of Microsoft 365 services that depended on Exchange Online and email were affected, while elements that did not rely on Exchange Online remained functional. Exchange Online is the company’s cloud-based messaging platform that delivers email, calendar, contacts, and tasks.

The company added that it was “not aware” of any malicious activity that was causing the outage.

Microsoft said it is “seeing positive results” for service stability and increased capacity as it continues to work on fixing the outage.

Earlier, reports suggested the problem was caused by an overloading of certain servers after others had failed.