Phishing and social engineering attacks increasingly arrive through chat apps rather than email, and Microsoft Teams is a growing target for impersonation attempts. For SMB administrators, understanding what Teams actually does to detect these messages—and what still depends on licensing or security tooling—matters when deciding whether additional protection is needed.
Microsoft’s own support documentation describes an AI-powered phishing detection capability built directly into Teams chat. A separate but related capability, described in a Microsoft Defender for Office 365 blog post, adds URL-based threat scanning and security operations tooling to Teams messages. These are two different protection mechanisms, and it’s worth understanding how they differ before assuming either one covers everything.
What the AI-powered phishing detection in Teams does
According to Microsoft’s support documentation, Microsoft Teams uses AI-powered, context-aware detection to help protect users from phishing and social engineering activity in chat conversations. Rather than simply scanning for known malicious links, this detection evaluates the context of a conversation to identify behaviors associated with phishing, such as patterns commonly seen when someone is trying to extract money or personal information.
Microsoft’s documentation is explicit that this detection is used only for security purposes. It states that phishing content detection isn’t used to train AI models, summarize conversations, generate productivity insights, monitor users, serve advertising, or support any other non-security purpose.
Availability and tenant scope
Per the supplied Microsoft documentation, this detection is available to users worldwide on mobile, desktop, and web versions of Teams. It applies to:
- One-on-one chats
- Group chats
The detection specifically looks at conversation context to identify phishing-related behavior coming from new external first-time contacts—in other words, people outside your organization who haven’t previously chatted with the user.
Expanded coverage for mobile users in China
Microsoft’s documentation notes an additional layer of protection for mobile users in China. For this group, phishing and fraud protection may extend to both external and in-tenant (internal) conversations, not just new external contacts. Microsoft describes this expanded coverage as designed to address region-specific fraud patterns. This distinction is documented only for mobile users in China; the supplied source does not extend this broader internal-conversation coverage to other regions or platforms.
How new-contact detection works
The documented mechanism centers on conversation context from first-time external senders. When someone outside the organization messages a user for the first time, Teams analyzes the conversation for phishing-related behavior. If the system flags a message as potentially harmful, a warning banner appears in the chat to help the recipient make a more informed decision before continuing.
Microsoft’s documentation does not describe additional technical detail about the detection model, confidence thresholds, or how “new” contact status is tracked over time. Administrators should treat this as a user-facing, in-chat warning system rather than a configurable security policy with adjustable rules, since no such admin-facing configuration is described in the supplied source.
What users are advised to do
Microsoft’s guidance for end users who see a phishing warning banner includes:
- Verify the sender’s identity — check the email address and message content for signs of impersonation, including requests for money or personal information.
- Block and report confirmed phishing — if phishing is confirmed, don’t engage with the sender; block and report them from their contact card.
- Correct false positives — if the warning was triggered for a trusted contact, use the “Report as not a concern” option from the flagged message’s options menu.
- Avoid sharing sensitive information in chat, including passwords, verification codes, and financial details.
These are documented Microsoft recommendations aimed at end users, and they’re worth reinforcing in SMB security awareness training.
Licensing and admin controls: what’s documented and what isn’t
This is an area where it’s important not to overstate what Microsoft has published. The supplied support article describing AI-powered phishing content detection does not state a specific license requirement, nor does it describe an admin-facing toggle, policy, or configuration setting for this particular capability. If your organization needs confirmation of licensing requirements or administrative controls specific to this feature, that detail isn’t present in the documentation reviewed for this article—check current Microsoft Teams admin documentation directly rather than assuming a particular tier applies.
A separate, licensed layer: Defender-based URL protection in Teams
Microsoft has also described a distinct set of threat protection capabilities for Teams messages containing URLs, delivered through Microsoft Defender. Unlike the AI conversation-based phishing detection above, this is a URL-scanning and security-operations capability with clearly documented licensing tiers:
- Teams enterprise license and above (worldwide): Near-real-time scanning of messages containing URLs, with warning tips shown to users when malicious links are detected. Protection continues for up to 48 hours after delivery in case a previously safe URL is later weaponized. This capability is described as enabled by default, with no additional configuration required.
- Microsoft Defender for Office 365 (MDO) Plan 2 or Microsoft 365 E5: Adds user-driven reporting (“Report as security risk” / “Report as not security risk”) directly from Teams messages, three Advanced Hunting tables (MessageEvents, MessagePostDeliveryEvents, MessageURLInfo) for security teams, the ability for security admins to remove users from unsafe chats via an Action Wizard, and the ability to block malicious domains directly from the Microsoft Defender portal into the Teams Admin Center blocked domains list.
- MDO Plan 1: Zero-hour auto-purge (ZAP) and Teams admin quarantine, which automatically move malicious URL-based messages in internal Teams chats and channels to admin quarantine for review. Microsoft describes this as enabled by default for customers on this plan.
It’s worth being clear that this Defender-based protection is a URL-scanning and SOC-response system, distinct from the AI conversation-context phishing detection aimed at new external contacts described earlier. The two systems address overlapping but not identical risks, and an SMB relying only on one should understand that the other may still apply depending on licensing.
Rollout timing and preview status
Based on the supplied Microsoft sources, the AI-powered phishing content detection support article was last updated in October 2026 and describes the capability as available to worldwide users now, without indicating a preview or opt-in status. The Defender blog post describing URL-based Teams protection was published earlier in 2026 and also does not describe a preview or opt-in requirement for the capabilities it covers.
Neither supplied source confirms a specific future rollout date, such as a November 2026 timeline, for any of these capabilities. Administrators should rely on current Microsoft Teams and Microsoft 365 admin center messaging or official roadmap communications for the most up-to-date rollout schedule rather than assuming a specific date not confirmed in this documentation.
Practical recommendations for SMB admins
These are practical suggestions, not documented Microsoft requirements:
- Review your Teams and Defender licensing mix. Confirm which tier of Teams and Microsoft Defender for Office 365 (if any) your organization holds, since the URL-scanning protections described above vary meaningfully by license.
- Reinforce user awareness around new external contacts. Since the AI phishing detection specifically targets first-time external senders, remind staff that unfamiliar contacts requesting money, credentials, or sensitive data warrant extra scrutiny even when no warning banner appears.
- Don’t assume full coverage from one feature. The AI conversation-based detection and the Defender URL-scanning capability serve different purposes; relying on only one may leave gaps depending on the nature of an attack.
- Monitor official Microsoft communications for rollout updates. Since the supplied sources don’t confirm a specific future rollout date, check the Microsoft 365 admin center message center or Teams release notes directly for the latest status.
Conclusion
Microsoft Teams includes a documented, AI-powered phishing detection capability that analyzes conversation context to flag risky behavior from new external contacts, available worldwide with expanded coverage for mobile users in China. Separately, Microsoft Defender for Office 365 adds URL-based scanning, quarantine, and security operations tooling to Teams messages, with protection levels tied to specific license tiers. For SMB administrators, the practical takeaway is to understand these as two complementary—but distinct—protections, confirm your current licensing position, and continue reinforcing basic phishing awareness among staff, since Microsoft’s own guidance still relies heavily on users recognizing and reporting suspicious messages themselves. As with any security capability still evolving, it’s worth checking official Microsoft documentation periodically for updates on scope, licensing, and administrative controls.