Real-Time Deepfake Detection for Microsoft Teams Meetings
In 2024, an Arup employee approved 15 transfers worth $25.6 million after a video call where his CFO and colleagues were all deepfakes. Nobody on that call was real except him. UncovAI now checks the face and the voice on every Teams call as it happens.
Fake faces and cloned voices now arrive together
The Arup case was not a one-off. In 2025, a Singapore finance director sent more than $499,000 after a call with a deepfake executive. In February 2026, a cloned voice helped move roughly €95 million out of Fideuram, part of Intesa Sanpaolo.
Gartner’s September 2026 survey of 297 CISOs found that 41% had dealt with a deepfake in a social-engineering attack on an employee phone call in the past year. Another 36% had seen one on a video call.
Attackers rarely fake just one channel. A swapped face comes with a cloned voice, and some feed synthetic video straight into the call instead of pointing a camera at a screen. Spotting that by eye, in the middle of a tense conversation, is a poor bet.
How UncovAI works in Microsoft Teams
Add the app from Microsoft Marketplace and a bot joins your Teams meetings automatically. No one starts detection by hand, and participants download nothing.
The bot analyzes the live video feed and the audio stream at the same time. It looks for face-swapping, voice cloning and synthetic stream injection, and reports a confidence score so your team can see how likely the media is to be fake. The same engines that power our video detection and audio detection run on the call.
When something looks wrong, the host decides what to do next. They can ask the person to confirm their identity another way, pause the request or bring in security.
Where it matters most
Any call that ends in a payment, a password reset, a hire or a signature is worth protecting.
CEO fraud
Flags a face-swapped or voice-cloned executive asking for an urgent transfer.
Remote hiring
Checks that the candidate on screen is a real person during interviews.
Payment approvals
Flags deepfaked finance leaders or vendors on calls about moving money.
Client and legal meetings
Verifies that clients are who they say they are before sensitive discussions.
For wider fraud and identity checks beyond meetings, see our AI scam and deepfake detector.
What to know before you deploy
The bot is a participant. It appears in the meeting while it analyzes the call. Some teams prefer that visibility, and it tells attackers someone is watching.
Scores are probabilities. Detection should sit next to other checks. For any high-value request, call back on a number you already trust.
It needs a stable connection. Analysis happens in real time over the internet.
You can try it first. A free trial includes limited daily credits. After that, usage is credit-based and needs an active subscription.
Zoom is covered too. The same protection is available as a Zoom app, so mixed teams get one standard.
Know who you’re talking to
SSO tells you which account joined the call. It cannot tell you whether the face and voice behind it are real. Real-time detection fills that gap.
Protect your meetings →