Particle.news

Incode Launches Deepsight to Spot Deepfakes in Under 100 Milliseconds

Independent testing by Purdue University ranked the detector best on accuracy with the lowest false-positive rate.

Overview

  • Deepsight uses a three-layer approach—Perception, Integrity, and Behavior—to tell real faces from AI output, validate device and camera authenticity, and flag micro-movement anomalies.
  • Incode says the system processes images and video in under 100 milliseconds to support real-time identity verification without adding user friction.
  • The tool operates within Incode’s identity platform without new hardware or infrastructure changes and works across devices for onboarding or authentication.
  • Incode reports rollouts with TikTok, PNC Bank, Scotiabank, and Nubank, claiming protection across more than 6 million live identity sessions.
  • The launch is positioned against rising digital fraud and AI-driven impersonation in Latin America, with cited research noting higher regional cyberattack rates and broad enterprise AI adoption.