This light-powered AI can spot deepfakes with nearly 98% accuracy
What happened
UCLA researchers developed an AI system that uses light-based analysis to detect deepfake videos. The system can scan more than a dozen videos at the same time and identifies deepfakes with nearly 98 percent accuracy. It operates quickly, consumes little energy, and is designed to resist adversarial attacks that often fool other AI detectors.
Why it matters
Deepfakes are flooding online platforms, creating real risks for misinformation, fraud, and reputational damage. Existing detection methods tend to be slow, require heavy computation, or can be tricked by evolving fakes. This new light-powered AI system reduces these bottlenecks by speeding up detection while lowering energy use. Its resistance to attacks makes it more reliable in hostile environments where bad actors try to bypass safeguards. This means platforms and businesses screening video content can deploy faster, cheaper, and harder-to-fool tools to maintain trust.
What to watch next
The key question is how this technology will be integrated into real-world systems. Operators should watch for open-source releases, APIs, or partnerships that bring this capability to content moderation platforms and verification services. Also important will be how it handles diverse video formats and scales beyond controlled testing. If UCLA’s AI can sustain accuracy and speed in the wild, it could pressure existing vendors to adopt light-efficient designs. Regulators and brand managers keeping an eye on video authenticity might push for adoption, increasing demand for tools that match this AI’s performance and resilience.
AI Quick Briefs Editorial Desk