Microsoft has mitigated an issue with one of its machine-learning (ML) models which incorrectly flagged Adobe email in Exchange Online as Spam.
The company revealed this in an advisory posted on the Microsoft 365 admin centre tagged as EX1061430users had problems accessing alerts regarding Adobe URLs beginning April 22 at 9:24 UTC, while being warned of a potentially malicious URL.
Usually, this type of alert is issued when Exchange Online users accidentally click on a link within an email. Microsoft said.
“To fix the issue we initiated Replay Time Travel (RTT) on the affected URLs to fully remediate impact. Impact was specific to some users who were served through the affected infrastructure.”
A final update posted on Thursday, April 24 at 11:04 UTC said that Microsoft had implemented mitigations in order to lower false positive rates. This was done by improving machine-learning algorithms to ensure that legitimate email messages won’t be incorrectly classified as spam, and therefore not delivered, in the future.
Although the company did not share any additional information about the regions or number of affected users, this incident was categorized as a service problem that is usually limited in scope and impact. Microsoft has dealt with similar issues in the past, resulting in emails being incorrectly classified as spam or quarantined. Microsoft, for example, addressed a false positive in Exchange Online that caused anti-spam systems incorrectly to quarantine emails from some users last month.
It also mitigated an Exchange Online Bug that tagged emails containing pictures as malicious and sent those emails to quarantine automatically. In October 2023, Microsoft had to disable a bad Anti-Spam rule that was flooding Microsoft 365 administrators’ inboxes by blind carbon copies of outbound emails that were mistakenly flagged spam.
