What are the fraud & abuse / trust & safety risks that an LLM based service is exposed to? How can we leverage LLMs to manage those risks? In this session we will go over the risks of exposing an LLM interface to your end-users. I will present my work on Base44 and show real examples of how those threats materialized. I will also present LLM based methods to defend against those risks. The session will focus on fraud & abuse / trust & safety risks and NOT on prompt injection / data leakage type of risks.
Room: Main hall
Tue, Oct 28th, 15:40 - 16:10