The field of membership inference attacks and synthetic data is rapidly evolving, with a focus on developing more effective methods for evaluating privacy risks and creating robust privacy-preserving mechanisms. Recent research has highlighted the vulnerability of generative models to membership inference attacks, emphasizing the need for caution when deploying these models in privacy-sensitive applications. Furthermore, studies have demonstrated the potential of synthetic data as a viable alternative to real data, particularly in areas such as facial recognition. Notably, synthetic datasets have been shown to achieve reliable recognition performance without compromising privacy. Additionally, there is a growing recognition of the connection between membership inference attacks and machine-generated text detection, with research exploring the transferability of methods between these two tasks. Overall, the field is moving towards developing more sophisticated and effective methods for protecting privacy and promoting the use of synthetic data. Noteworthy papers include: The Hidden Cost of Modeling P(X) which demonstrates the heightened susceptibility of generative classifiers to membership inference attacks. OpenLVLM-MIA which introduces a controlled benchmark for evaluating membership inference attacks on large vision-language models. Beyond Real Faces which presents a comprehensive empirical assessment of synthetic facial recognition datasets and their potential to replace real datasets. Machine Text Detectors are Membership Inference Attacks which investigates the transferability of methods between membership inference attacks and machine-generated text detection.