The field of AI-generated image detection is rapidly progressing, with a focus on improving robustness to real-world degradations such as motion blur. Researchers are exploring innovative methods to enhance the performance of AI-generated image detectors, including knowledge distillation and gradient surgery. These approaches aim to preserve the generalization ability of pre-trained models while adapting to new tasks and environments. Notably, the development of robustness evaluation frameworks is also underway, enabling fine-grained analysis of appearance-based gait recognition systems and other applications. Noteworthy papers include: DINO-Detect, which achieves state-of-the-art performance in blur-robust AI-generated image detection. RobustGait, a framework for evaluating the robustness of appearance-based gait recognition systems. DGS-Net, a novel framework that preserves transferable pre-trained priors while suppressing task-irrelevant components in CLIP fine-tuning. Dataset Distillation for Pre-Trained Self-Supervised Vision Models, which introduces a method for distilling datasets that enable optimal training of linear probes on top of large, pre-trained vision models.