The field of media forensics and safety is moving towards developing more sophisticated methods for detecting and localizing manipulated media, such as images and videos. Researchers are exploring new approaches, including context-aware contrastive learning and semantic-augment erasing, to improve the accuracy and generalizability of media forensic techniques. Additionally, there is a growing focus on ensuring the safety of generative models, particularly diffusion models, by developing methods to erase unacceptable concepts and prevent copyright infringement. Noteworthy papers include: RADAR, which proposes a novel approach for reliable identification of diffusion-based image manipulations, and SAGE, which introduces semantic-augment erasing to explore the boundaries of unsafe concept domains. These innovative techniques are advancing the field and have significant implications for real-world applications, such as image and video authentication, and safe content generation.