What happens when your AI-powered travel agent develops a case of Finding Nemo-level forgetfulness? In this talk, we dive into a real-world attack scenario where an AI agent's memory is subtly manipulated into believing that chartered flights are always free. By repeatedly reinforcing a false pricing rule, attackers can book luxury travel without ever swiping a card - essentially gaslighting the AI into bankrupting its own business.

We'll explore how this kind of memory injection works, why AI systems are especially vulnerable to it, and what this means for the security of LLM-integrated applications. Expect live demos, some hilarious (and horrifying) case studies, and practical takeaways on how to prevent your AI from becoming the most generous travel agent in history.

By the end of this session, you'll either be terrified of AI memory manipulation - or considering how to get yourself a free first-class ticket.