Microsoft has successfully resolved a machine learning issue that caused legitimate Gmail messages to be incorrectly flagged as spam in Exchange Online, the company confirmed this week.
The problem, tracked internally as EX1064599, began affecting users on April 25 and resulted in emails from Gmail accounts being automatically redirected to junk folders.
Microsoft engineers identified that their machine learning model, designed to protect Exchange Online users from malicious emails, was incorrectly categorizing legitimate messages due to pattern similarities with known spam.
“We’ve identified that our machine learning model, which safeguards Exchange Online against risky email messages, is incorrectly identifying legitimate email messages as spam due to their similarity to email messages used in spam attacks,” Microsoft explained when acknowledging the issue in the Microsoft 365 admin center.
After nearly a week of investigation, Microsoft addressed the problem by reverting to a previous, stable version of the machine learning model.
The fix was completed on May 1, with confirmation that service had been fully restored.
“After a period of monitoring, we’ve confirmed through our service health telemetry that the completion of reverting to the previous ML model has successfully remediated impact,” Microsoft stated in its final update on the incident.
During the outage, system administrators had the option to implement custom allow rules as a temporary workaround to ensure Gmail messages reached their intended recipients rather than being diverted to junk folders.
The company has not disclosed how many customers or which regions were affected by the issue, though the classification as an “incident” suggests the impact was significant enough to disrupt normal service operations for many users.
This isn’t the first time Microsoft has encountered false positive issues with its email security systems.
Just last week, the company resolved a similar problem where its machine learning algorithms incorrectly flagged Adobe emails as spam.
In March, another Exchange Online false positive incident resulted in legitimate emails being incorrectly quarantined.
Microsoft acknowledged the recurring nature of these issues, stating they are “continuing to investigate opportunities to improve our ML detection process to reduce false positive detections and prevent similar future impact.”
The pattern of email filtering issues extends back to October 2023, when Microsoft had to disable a problematic anti-spam rule that was flooding Microsoft 365 administrators’ inboxes with blind carbon copies of outbound emails mistakenly identified as spam.
More recently, in August 2024, the company addressed another Exchange Online bug that was incorrectly quarantining emails containing images after tagging them as malicious.
You must be logged in to post a comment.