The field of large language models (LLMs) is rapidly evolving, with significant advances in personalized user experience and task execution. However, this growth also introduces new security vulnerabilities and challenges. Researchers are actively exploring the safety risks associated with LLMs, including the potential for manipulation and misinformation. The use of LLMs in research also raises concerns about scientific integrity and the need for transparent prompt documentation. Noteworthy papers in this area include the proposal of CheatAgent, a novel attack framework that harnesses the capabilities of LLMs to attack LLM-empowered recommender systems. Another significant contribution is the development of CrossInject, a framework that exploits cross-modal prompt injection attacks in multimodal agents. The paper on Prompt-Hacking highlights the importance of critically evaluating the use of LLMs in research and advocating for transparent prompt documentation. Finally, the case study on Breaking the Prompt Wall demonstrates the vulnerability of commercial-grade LLMs to subtle manipulations and stresses the need for developers to prioritize prompt-level security.