Microsoft will add AI phishing detection to Teams that looks beyond links

A suspicious Teams conversation would not need a malicious link or attachment to draw scrutiny under Microsoft’s planned AI protection. The company has disclosed context-aware phishing detection scheduled to begin rolling out in November 2026. The feature remains in development, rather than being available to users now.

The conversation becomes a security signal

Microsoft says the feature will analyze conversation context for phishing, social engineering, impersonation attempts and deceptive intent. Social engineering means manipulating someone into cooperating with a scam; impersonation means pretending to be someone else. The surrounding exchange becomes part of what the detector evaluates.

Higher-risk situations include external first-contact chats, where someone outside an organization approaches a user for the first time. These are among the scenarios Microsoft identifies for the new protection.

TechRadar describes the capability through a Microsoft 365 roadmap entry and a Message Center update. It notes that the Microsoft post does not specify which AI platform powers detection.

Rollout does not mean immediate warnings

TechRadar describes a worldwide November rollout. Computerworld gives a more specific timetable: rollout beginning in early November 2026, with general availability expected in the first half of December. November is the planned starting window, not a promise that deployment will be complete that month.

In its Message Center update, quoted by Computerworld, Microsoft says initial quality validation may happen in the background without displaying warnings. After validation, users may see a warning when the system detects potential phishing.

The feature will be automatically enabled, with no configuration required from IT teams.

This capability is used only for security protection and not for productivity, summarization, advertising, user monitoring, or other non-security purposes.

Microsoft, in a Message Center update quoted by TechRadar

Microsoft also says customer message content will not be retained by the detection system for model training or other secondary uses. That commitment specifically limits the detection system’s retention and reuse of message content.