Microsoft Unveils Deepfake Detection Tool Ahead of U.S. Election
Microsoft releases Video Authenticator, a new tool that generates a confidence score to help identify artificially manipulated media. The company warns the technology only offers temporary utility against AI-generated deepfakes.
Microsoft introduces Video Authenticator, a new tool designed to analyze videos and photos to determine if they are artificially manipulated. The software provides a real-time confidence score for each frame of a video by detecting blending boundaries and subtle fading elements that the human eye easily misses.
The tech giant acknowledges this detection method only offers a temporary solution in the ongoing disinformation arms race. Because the artificial intelligence behind deepfakes constantly learns and evolves, Microsoft admits these manipulated videos will inevitably bypass conventional detection technologies in the long run.
Despite this looming limitation, Microsoft partners with the AI Foundation to provide the tool to organizations involved in the democratic process for the upcoming U.S. election. The company trains the system using leading public datasets, though it emphasizes that a critically thinking mind remains the absolute best defense against high-tech disinformation.