With the growth of the metaverse and XR technologies, BCIs are expanding from medical applications to various consumer industries. However, this broader adoption has raised concerns about the privacy and security of EEG data. To address this challenge, we propose the RPW-EEG framework, which embeds copyright information as perturbations to enhance the security and traceability of data, while ensuring the robustness and usability of the data. The framework adopts an encoder-decoder architecture for end-to-end training, incorporating a noise layer to enhance the stability and anti-attack capabilities of the watermark data. Additionally, to prevent the loss of task-related features in EEG data, we introduce a plug-and-play fine-tuning module that restores these features within the watermark-embedded signals. Experimental results show that RPW-EEG outperforms baseline models in watermark quality, with extraction accuracy exceeding 61% under various attacks, and achieves 88.5% classification accuracy for task paradigms, effectively balancing copyright protection and EEG data usability.