In today’s Cloud Wars Agent and Copilot Minute, I explore how Microsoft’s new AI-powered phishing detection in Teams could transform cybersecurity while raising important questions about user privacy.
Highlights
00:10— Microsoft is introducing a new AI feature to Teams that will detect phishing attempts and scams by analyzing conversations. Now, rather than relying solely on traditional methods of detecting malicious links and files, it uses behavior-based detection to identify suspicious activity in messages.
00:28 — The feature will initially run silently in the background, with warnings potentially appearing when phishing is detected. It will also prioritize conversations with new contacts, as the risk here is generally higher. This is a great use of AI. It shows how companies can use the technology to approach problems differently. In this case, by analyzing behavior at scale to strengthen a company’s cybersecurity defenses.
01:03 — Microsoft says message content won’t be retained for AI training or other secondary uses, and it says the feature will only be used for security purposes. Now, the rollout is due to begin in November 2026, with general availability expected in the first half of December this year. Crucially, it will be enabled automatically, without requiring IT administrators to configure it.
01:31 — And that’s where, for me, things get interesting, complicated. Think back to Microsoft Recall. That’s that controversial AI feature that took screenshots of users’ screens and was initially planned to be enabled by default. Now, following some pretty significant privacy and security concerns, Microsoft went back to the drawing board and made saving screenshots on Microsoft Recall an opt-in feature.
01:51 — So, introducing another automatically enabled AI feature that analyzes, again, conversations raises an obvious question: Has Microsoft learned from its past mistakes, or is it repeating them?
