CRESTCon Europe 2024 - Should You Let ChatGPT Control Your Browser - Donato Capitella

This presentation will explore the practical risks of granting LLMs agency, that is allowing them to perform actions on a user’s behalf, and how
attackers might be able to exploit this in the real world.
It will specifically look at an emerging use case - autonomous browser agents - and will explore how LLM agents work, the risk of indirect prompt injection and how to mitigate this.