AI Deepfake Real-Time Detection Platform VerifyX Launches: Browser Extension Tags 95% of AI-Generated Images

Full Article

Stanford's AI Safety Lab officially launched the VerifyX deepfake detection platform on May 11. The platform provides Chrome and Firefox browser extensions along with RESTful API interfaces, capable of detecting and tagging AI-generated images, videos, and audio content in real-time as users browse the web.

VerifyX's core detection engine is based on a multimodal model trained on 20 million annotated images. The model analyzes not only image pixel statistics but also combines C2PA metadata verification, reverse image search, and generative model fingerprint identification as auxiliary detection methods. On standard test sets, VerifyX achieved 95.2% accuracy for AI-generated images, 91.7% for AI-generated videos, and a false positive rate of 2.3%.

Stanford AI Safety Lab director Dan Boneh said: "Deepfakes are no longer a future threat but a present reality. VerifyX aims to give ordinary users a simple, accessible tool to distinguish real from fake while browsing the web."

The VerifyX extension surpassed 1 million downloads within 48 hours of appearing on the Chrome Web Store. However, the arms race nature of detection technology means VerifyX's accuracy may decline over time—as generative models learn to evade detection features, detection models must also update accordingly.