One paper for protecting multimodal data from unauthorized VLM fine-tuning has been released on arXiv.