📰 Key Takeaways

Zoom recently announced that enterprise users will be able to verify that the person on the other end of a video call isn’t an AI deepfake through a “Proof of Human” badge. The biometric technology was developed by World, the company backed by Sam Altman, and the system is called “World ID for Zoom,” now live globally on the Zoom Marketplace. Here’s how it works: the system cross-checks three pieces of data — the iris scan captured by World’s Orb device, a live selfie, and the video feed from the meeting. Once all three match, a “Human Badge” appears on screen, visible to everyone else in the call. The feature was tested in a small pilot with VanEck Funds back in April, and has now rolled out fully to enterprise customers. World also noted that financial institutions like Bain Capital and Pantera Capital were early design partners on its identity verification technology. This launch comes on the heels of several AI impersonation scams in recent years: in 2024, an employee at engineering firm Arup was tricked into wiring about $25 million after scammers impersonated executives during a video call. This past February, hacking groups believed to be linked to North Korea used fake Zoom video feeds to trick employees at crypto and fintech companies into installing malware. World says the technology is designed to protect highly sensitive, high-risk conversations by ensuring the person on screen is a real, unique human being.

💬 JudyAI Lab Take

Zoom’s partnership with World to launch a “Proof of Human” badge — using iris scans to verify video feeds and block AI deepfake impersonation — shows just how far deepfake scams have moved from theory into real corporate boardrooms.

There’s a design shift worth watching here: as AI-generated content gets more convincing, identity verification is moving away from “what password do you know” toward “are you actually a real, existing human” — swapping biometrics in for traditional authentication. The Arup employee who got scammed out of $25 million by someone impersonating an executive, the North Korea-linked hackers faking video feeds to plant malware — both cases show the social engineering attack surface has shifted to live audio/video, not just text or email. For AI builders, this is a reminder: if your product involves any kind of identity judgment, “is the other party actually human” needs to be treated as its own risk category, not something you just eyeball.

If your product involves video, voice, or real-time interaction, it’s worth checking: for high-risk moments like wire transfers or permission changes, do you have an extra layer of identity confirmation in place?

📅 Source Info

🔗 Further Reading