Artificial intelligence and deep learning allow for both opportunities and challenges in intellectual property protection. The technical tools of artificial intelligence and deep learning can be used for the protection and management of intellectual property rights to improve efficiency and reduce costs. The purpose of this paper is to propose a watermark extraction and embedding technique based on Transformer to facilitate the protection of intellectual property rights. This paper adopts the literature review method and theoretical analysis method to construct an intellectual property protection model based on AI and deep learning, which can realize the embedding and extraction of watermarking information in deep learning models, so as to realize the ownership authentication and traceability of deep learning models. In this paper, an experimental analysis was carried out through public datasets to compare the model with other similar models. The results show that the model can accurately extract the watermark information and is robust against various attacks. The Transformer-based watermark embedding and extraction model is an effective video watermarking method that can effectively protect the intellectual property rights of videos.