As businesses rush to embrace the perceived benefits of AI systems, security professionals must take a more pragmatic view. In this talk, Andy will highlight some of the surprising attack vectors that LLM-powered applications may vulnerable to - and what we can do to help prevent abuse.