ProMark is a novel proactive watermarking technique designed to attribute synthetic images to their training data concepts, such as objects, motifs, templates, artists, or styles. The method embeds imperceptible watermarks into the input training images and trains diffusion models to retain these watermarks in the generated images. ProMark can embed up to \(2^{16}\) unique watermarks into training data, allowing each image to contain multiple watermarks. The approach maintains image quality while outperforming correlation-based attribution methods. ProMark is evaluated on five diverse datasets (Adobe Stock, ImageNet, LSUN, Wikart, and BAM) and demonstrates higher accuracy in attributing generated images to their training concepts. The method is flexible, applicable to both unconditional and conditional diffusion models, and can be fine-tuned with minimal changes. ProMark provides a causal attribution framework that offers opportunities for recognizing and rewarding creative contributions to generative AI, contributing to the future creative economy.ProMark is a novel proactive watermarking technique designed to attribute synthetic images to their training data concepts, such as objects, motifs, templates, artists, or styles. The method embeds imperceptible watermarks into the input training images and trains diffusion models to retain these watermarks in the generated images. ProMark can embed up to \(2^{16}\) unique watermarks into training data, allowing each image to contain multiple watermarks. The approach maintains image quality while outperforming correlation-based attribution methods. ProMark is evaluated on five diverse datasets (Adobe Stock, ImageNet, LSUN, Wikart, and BAM) and demonstrates higher accuracy in attributing generated images to their training concepts. The method is flexible, applicable to both unconditional and conditional diffusion models, and can be fine-tuned with minimal changes. ProMark provides a causal attribution framework that offers opportunities for recognizing and rewarding creative contributions to generative AI, contributing to the future creative economy.