Researchers from the University of California, Los Angeles (UCLA) have created a new type of artificial intelligence system that uses light to identify deepfake videos. This hybrid system combines traditional computing with optical processing, allowing it to analyze more than fifteen video streams at the same time. In laboratory tests, it achieved an accuracy of 97.79 percent and a sensitivity of 99.86 percent, meaning it was very good at correctly identifying both real and fake videos. The system works by using a lightweight numerical encoder to extract key features from each video, such as its spatial, spectral, and temporal characteristics. These features are then transformed into a phase pattern and displayed on a programmable spatial light modulator. The optical wavefront created by this process passes through a passive free-space optical decoder. At the end of this process, paired optical detectors assign each video an authenticity score, determining whether it is real or fake. The researchers tested the system on the Celeb-DF dataset, a widely used collection of deepfake videos. The system achieved an average accuracy of 97.79 percent when analyzing 15 videos at once. When the number of videos increased to 18, the accuracy slightly dropped to 96.13 percent. The system was also tested on videos created using VEO-3, an AI model developed by Google. After adjusting the model with videos generated by Google's Gemini AI, the system achieved an accuracy of 94.80 percent, with a sensitivity of 97.61 percent and a specificity of 92.00 percent. To improve performance, the researchers added passive optical layers to the system. These layers, which are static and do not require electrical power, helped increase detection accuracy on the DeepSpeak dataset from 88.98 percent to 95.76 percent. The system was also tested under various degradations, such as noise, blur, and compression. It maintained its detection capabilities under these conditions, showing that it could be a reliable tool as deepfake technology continues to evolve.