This paper addresses the ethical concern of protecting AI-generated content, particularly in the context of diffusion models. The authors propose a unified watermarking framework called WaDiff, which integrates user-specific watermarks into the generation process of diffusion models. WaDiff aims to enhance both detection and owner identification of AI-generated content. The framework is designed to be scalable and efficient, without requiring customized fine-tuning, and it maintains the quality of generated images while embedding unique fingerprints. Extensive experiments on popular diffusion models demonstrate the effectiveness and robustness of WaDiff in detecting and identifying AI-generated content, outperforming existing post-hoc watermarking strategies. The method is publicly available, and the authors provide detailed implementation details and ablation studies to support their claims.This paper addresses the ethical concern of protecting AI-generated content, particularly in the context of diffusion models. The authors propose a unified watermarking framework called WaDiff, which integrates user-specific watermarks into the generation process of diffusion models. WaDiff aims to enhance both detection and owner identification of AI-generated content. The framework is designed to be scalable and efficient, without requiring customized fine-tuning, and it maintains the quality of generated images while embedding unique fingerprints. Extensive experiments on popular diffusion models demonstrate the effectiveness and robustness of WaDiff in detecting and identifying AI-generated content, outperforming existing post-hoc watermarking strategies. The method is publicly available, and the authors provide detailed implementation details and ablation studies to support their claims.