A Friday-the-13th technology pile-up unfolded overnight as a Microsoft Exchange Online outage disrupted Outlook and Microsoft 365 connectors, triggering intermittent failures in Anthropic’s Claude and OpenAI’s ChatGPT – although Microsoft ruled out a cyber attack.
The issues started around 6pm SAST last night when users began experiencing problems accessing their mailboxes through Outlook on the web, as well as desktop and mobile apps, while some e-mail messages were also delayed.
Microsoft subsequently reported a wider problem affecting access to multiple Microsoft 365 services.
Tom’s guide, at 6.30am, was saying that “something is seriously wrong with Microsoft Outlook and other popular Microsoft services like Teams, Microsoft 365, Copilot and others, as more than 100 000 users are flocking to the popular outage-reporting website Downdetector to report issues”.
The website adds that this number has “dropped a bit since then, but it’s still a very widespread outage”, while Downdetector, which counts user-submitted problem reports for each service rather than users, showed 908 reports for Microsoft 365 as of 9am.
Traffic jam
Microsoft says the outage persisted for more than 12 hours because traffic was redirected away from servers that had become unavailable, placing additional pressure on the remaining operational infrastructure. The increased load overwhelmed those servers, which also began intermittently failing to process requests and complicated Microsoft’s efforts to restore the service.
Microsoft had redirected some traffic away from affected infrastructure in the US to reduce pressure, identified the servers that were intermittently becoming unavailable, and then started to redistribute Outlook traffic across its global network, says the University of Pennsylvania’s Penn Computing.
The software company says it is pursuing multiple workstreams to pinpoint the underlying cause of the traffic influx and advance the recoverability of the backend servers. “We’re taking a measured approach to identify and repair affected infrastructure.”
Microsoft initially reported positive results from restarting affected infrastructure, but at 5am it said the changes intended to improve service availability were not working as expected and that it was looking at alternative ways to increase capacity, while dealing with imbalances in how traffic was being managed.
“As of now, we are not aware of any malicious activity that is causing this incident. We will continue to treat this incident with urgency and priority until full resolution,” Microsoft said in its latest update as of the time of writing at 8am.
Microsoft adds that, “once mitigated, we will provide full details of the root cause, defence-in-depth repair items, and more within a Post-Incident Report”.
Cloudy outlook
Linda Morris, director of Smart Technology Centre, which hosts Microsoft Exchange accounts, says the latest Microsoft connectivity disruptions across multiple regions highlight the vulnerability of hosted services to downtime. She says its Microsoft servers were sporadically down until around 3am and fully online as of 8am.
In an interconnected business environment, modern multi-factor authentication and single authentication layers mean localised failures can quickly disrupt workflows, says Morris.
According to a Claude Status page, the intermittent downtime has been caused by a Microsoft 365 service incident with mail, calendar and search tools in the Microsoft 365 connector failing for most requests as of midnight. “Claude for Outlook remains affected… We will update this post when service is restored,” it says in its latest post.
OpenAI states it is investigating an issue with its Microsoft connectors, warning that new connections may fail and existing tools may not work as expected. The incident, which affects ChatGPT’s Connectors/Apps component, had been ongoing for 11 hours as of the time of checking on Friday morning.
The artificial intelligence company has not published a specific uptime percentage for the incident, noting its availability metrics are reported at an aggregate level across subscription tiers, models and error types. Independent monitoring by Openuptime puts OpenAI’s estimated uptime at 98.31% over the past 90 days, based on 20 reported incidents, including two full outages.
“What makes this outage particularly significant is not simply that Microsoft 365 experienced disruption, but how quickly the impact extended into the broader technology ecosystem, including AI platforms such as ChatGPT and Claude,” says Jacqui Muller, a researcher at Belgium Campus iTversity and a PhD candidate in computer science and IT.
Pen and paper
“When the cloud goes down and AI isn’t available to draft your e-mails, someone still has to know how to issue a manual invoice or draw on a whiteboard,” says Morris.
The latest incident is not the first Microsoft 365 disruption in recent weeks. Microsoft’s incident register records a series of separate problems dating back to 6 October, affecting Outlook and Teams contacts, e-mail delivery, calendars, Teams meetings and calls, Copilot, SharePoint, OneDrive and administrative tools.
Microsoft’s reported causes included faulty software updates, configuration changes, network infrastructure problems and backend processing failures. While most were subsequently resolved, some remained open at the time of the latest report.
“Many businesses mistakenly assume cloud providers take 100% responsibility for data protection; in reality, while infrastructure uptime is guaranteed, data backup and recovery remain the customer’s ultimate responsibility,” Morris adds.
Morris says maintaining a mirrored secondary data centre is often prohibitively expensive, while switching to another facility is complicated by dependencies such as the Domain Name System. “Cloud services are amazing, but no hosted service is immune.”
* All times are South African Standard Time (SAST), which is two hours ahead of UTC, the time zone used in Microsoft's incident reporting.
Share
