The field of retrieval-augmented generation is moving towards addressing the security and privacy vulnerabilities introduced by the integration of external knowledge bases. Researchers are exploring innovative methods to detect and prevent attacks, such as corpus poisoning and knowledge extraction. One notable direction is the development of context analysis techniques that can identify malicious content without relying on internal model knowledge. Another area of focus is the design of flexible and adaptable solutions for content moderation, which can quickly respond to emerging threats. The importance of privacy-preserving techniques is also being highlighted, particularly in the context of multimodal retrieval-augmented generation. Noteworthy papers in this area include: EcoSafeRAG, which achieves state-of-the-art security with plug-and-play deployment and improves clean-scenario performance. RAR, which offers superior flexibility and real-time customization capabilities for content moderation. Beyond Text, which reveals privacy vulnerabilities in multimodal retrieval-augmented generation and highlights the need for robust privacy-preserving techniques. Silent Leaks, which introduces an implicit knowledge extraction attack that can extract private information through benign queries. Chain-of-Thought Poisoning Attacks, which proposes a novel approach to attacking RAG systems by simulating chain-of-thought patterns aligned with the model's training signals.