UCLA's Optical-Neural Processor Screens 15+ Video Streams Simultaneously
Researchers at the University of California, Los Angeles (UCLA) have developed a new optical-neural processor that uses light to identify deepfake videos quickly and accurately. Unlike conventional systems that typically examine videos one at a time using digital hardware, the UCLA technology can analyze 15 or more video streams simultaneously.
The key difference lies in how the detection process is carried out: part of it takes place through the physical propagation of light. This allows many videos to be evaluated during a single optical pass, rather than requiring each one to move separately through a conventional digital processing pipeline.
The work is detailed in the study "Scalable, Energy-Efficient Optical-Neural Architecture for Multiplexed Deepfake Video Detection," published in eLight. The researchers designed the optical AI system to serve as a high-throughput, attack-resilient first layer of defense for screening large volumes of manipulated and AI-generated video.
The Growing Challenge of Deepfake Detection
Rapid improvements in generative AI have made synthetic videos increasingly realistic, heightening the need for detection systems that are both accurate and capable of operating at large scale.
Many advanced deepfake detectors depend on enormous amounts of digital computation. A single analysis can require hundreds of billions of floating-point operations, and videos are often processed sequentially. As more content must be checked, both processing time and energy requirements can rise proportionally.
Digital detection systems face another problem: attackers can deliberately alter fake videos in subtle ways designed to confuse a detector and make manipulated footage appear authentic.
Professor Aydogan Ozcan and his UCLA team developed a hybrid digital-optical system intended to address both limitations. By combining optical computing with digital processing, the system aims to deliver high accuracy, low energy consumption, and resistance to adversarial attacks.
How the Optical AI Works
In the UCLA architecture, video frames are encoded onto light beams that pass through a series of optical elements. These elements perform parts of the neural-network computation in the analog domain, exploiting the inherent parallelism of optics. Because light beams can travel through the same optical components independently, multiple video streams can be processed at once without interfering with one another.
The optical front end handles feature extraction and early detection stages, while a digital backend completes the classification. This division of labor allows the system to maintain high accuracy while dramatically reducing the computational burden on digital hardware.
According to the researchers, the system achieves nearly 98% accuracy in detecting deepfakes across multiple benchmark datasets. It also demonstrates resilience against common adversarial attacks, which attempt to fool detectors by adding imperceptible noise or perturbations to video frames.
Why It Matters in 2026
As generative AI tools become more accessible and capable, the volume of AI-generated video is exploding. In 2026, deepfakes are no longer a niche concern: they are used in misinformation campaigns, financial fraud, and non-consensual imagery, among other harms. Platforms, newsrooms, and regulators are under increasing pressure to detect and flag synthetic media at scale.
Traditional digital detectors struggle to keep pace. Processing videos one by one is too slow and too energy-intensive for real-time screening of the vast quantities of content uploaded daily. Optical computing offers a potential path forward by performing many operations in parallel with light, which is inherently faster and more energy-efficient for certain tasks.
The UCLA system's ability to analyze more than a dozen video streams simultaneously could make it suitable for deployment at the edge of large platforms or within content-moderation pipelines. Its low energy demands also make it attractive for continuous operation.
Limitations and Next Steps
The research represents a proof of concept, and several challenges remain before widespread deployment. Fabricating optical components at scale, integrating the system with existing digital infrastructure, and ensuring robustness across diverse deepfake generation methods are all active areas of investigation.
Nevertheless, the work highlights a promising direction: using the physics of light to tackle one of the most pressing problems in AI safety. As deepfakes grow more sophisticated, hybrid optical-digital approaches may become an essential part of the defense toolkit.
The study was published in eLight, an open-access journal focused on optics and photonics. The research team includes members from the UCLA Samueli School of Engineering and the California NanoSystems Institute.