Arizona State University professor Dr. YZ Yang has developed a "decentralized attribution technique" developed by his team to detect AI deepfakes, using a digital fingerprint system that can trace AI-generated content back to its source. The technique involves slightly altering the weights or internal settings within AI models like image generators and chatbots to create a unique signature in every generated image, hidden in clusters of pixels. The technology could potentially increase the reliability of content generated by AI, but rogue actors could still release open-source models without the watermarking protocol, leaving a potential gap in coverage. Yang's research also includes a system that could predict traffic crashes before they occur, using computer vision linked to existing traffic cameras to help drivers and pedestrians.