Zoom Plans Photorealistic AI Avatars Amid Growing Deepfake Concerns
Zoom announces a new feature that creates AI-animated, photorealistic avatars from short video clips to help users send asynchronous messages. The tool raises significant security questions as the company remains vague about specific safeguards against deepfake misuse.
Zoom announces a new feature that creates photorealistic AI avatars from short video clips to help users send asynchronous messages. The tool generates a digital clone of a user's head, upper arms, and shoulders, allowing them to type a script that the avatar speaks with synced lip movements. Chief product officer Smita Hashim states that these avatars save time and enable users to scale their video creation efforts.
The technology raises significant deepfake concerns as malicious actors increasingly use digital clones to spread disinformation. While competitors like Tavus and Microsoft implement strict safeguards such as verbal consent statements and written permission requirements, Zoom takes a vaguer approach. Hashim points to existing usage policies and mentions "advanced authentication" and watermarking as upcoming safety measures, but lacks specific details on how the system prevents impersonation.
This new feature aligns with CEO Eric Yuan's broader vision of creating AI stand-ins that eventually attend meetings, answer emails, and make phone calls on a user's behalf. However, the rollout comes at a precarious time when deepfakes of public figures spread rapidly across social media and blur the line between truth and fabricated media. Zoom plans to release the custom avatar feature sometime next year.