1 paper
Zhihao Wu, Gracia Gong, Qinglin Zhu +2
Watermarking embeds statistical signatures in AI-generated text for detection and attribution. We reveal a fundamental vulnerability: when users access multiple models (today's rea…