Identity verification startup Incode Technologies has launched Deepsight, a new multimodal AI system designed to detect and defend against AI-generated deepfakes used to impersonate real people during video calls, a threat that has surged 148% this year, SiliconANGLE reports.
The system employs a three-layered defense, analyzing video, motion, and depth data simultaneously to spot subtle anomalies and verify the authenticity of the camera and device being used. According to Incode founder and CEO Ricardo Amper, deepfakes have evolved from novelties into major fraud weapons, and when "identity can be faked, everything breaks."
The company claims Deepsight not only identifies AI-generated callers but can also pinpoint the specific generative AI model used to create the fake. Its effectiveness is backed by a Purdue University study, where it achieved the highest accuracy and lowest false acceptance rate among 24 commercial and academic detection tools. Incode's internal tests also suggest the system is ten times more accurate than trained human reviewers.
