Zoom integrates Verified Human from World to protect video meetings from deepfakes

Zoom announced a partnership with World — a biometric project linked to Sam Altman — and is adding to Meetings a Verified Human mechanism designed to confirm that the person on your video call is a real person and not a generated deepfake. The solution targets not regular standups, but sensitive calls where an error could cost millions.

How Deep Face Works

The integration is built around World's World ID Deep Face technology. During verification, the system compares three signals: a frame from the live video stream in Zoom, a Face Auth selfie verification on the participant's device, and an image obtained during initial registration through Orb — World's proprietary biometric scanner. If everything matches, a Verified Human badge appears next to the user's name and video. According to the companies, verification is performed locally on the device, and Zoom only receives the verification result, without transmitting personal biometric data to other meeting participants.

• Verification on meeting entry
• Deep Face Waiting Room before admission to the call
• Request for re-verification during the call itself
• Verified Human badge in the meeting interface

But this is not a mass-market feature in the "set and forget" spirit. To pass such verification, a person needs the World App and a pre-confirmed World ID obtained through a physical Orb device. World has around 18 million verified users across 160 countries and approximately 1500 active Orbs. For Zoom with its massive user base, this is for now more of a narrow trust layer for specific scenarios, rather than a universal standard for all work calls.

Why Zoom Is Moving Into Biometrics

The reason is simple: deepfake fraud has stopped being a rare experiment. In the first quarter of 2025, businesses lost over $200 million to such attacks. One of the most high-profile cases was the engineering company Arup, which lost $25 million in 2024 after a video call where all participants, including an allegedly chief financial officer, turned out to be AI-generated doubles. In 2025, a similar attack affected an international company in Singapore.

There are already tools on the market that try to catch fakes by analyzing the video frames themselves. But Zoom and World are betting on a different approach: rather than trying to guess whether the screen was drawn by a neural network, they confirm identity through a pre-created biometric record. This aligns well with Zoom's strategy as a corporate platform that needs to sell not just convenience and AI features, but trust for negotiations, payment approvals, executive communications, and other high-risk scenarios.

"Security and trust have always been the foundation of our platform," —

Brendan Ittelson, Zoom's Director of Ecosystem.

Where the Risks Begin

This approach comes with a cost. World has long faced scrutiny from regulators and privacy advocates over biometric collection, particularly iris scans. In several jurisdictions, including Spain, Germany, and the Philippines, the project has already faced complaints. World responds that it uses a privacy-focused architecture, zero-knowledge proofs, and local storage of sensitive data, but for many companies this will not be enough: the very fact of mandatory biometric registration could already become a legal and reputational problem.

There is also a purely practical barrier. Deep Face is currently in limited beta, and only Orb-verified World ID holders can pass verification. This means that for routine meetings with contractors, clients, or candidates, the feature is too cumbersome. However, for treasury teams, M&A negotiations, telemedicine, legal approvals, and other calls where the cost of error is very high, such friction could prove well justified. It is also important to note that the integration analyzes video, not voice.

What This Means

Protection against deepfakes in video communications is shifting from content detection to identity verification. Zoom shows that for business, the question "is this definitely a person?" is becoming a separate layer of security. But mass adoption of such systems will stumble not on model quality, but on whether companies and users are willing to accept biometrics as the price for trust.