Breaching LLM-Powered Applications: Overcoming Security and Privacy Challenges by Brian Vermeer

This talk explores the intersection of large language models (LLMs) and security, focusing on the risks associated with AI applications. The speaker addresses critical challenges such as prompt injection, key leakage, and the misuse of customer data for training models, while also considering legal implications. Attendees will gain insights into the vulnerabilities to avoid, strategies for ensuring data privacy compliance, and best practices for developing secure applications powered by LLMs. The session emphasizes the importance of general security in system design and its influence on LLM behavior and outcomes.