
Hacking AI Agents: Application Instructions for LLM and Tool Vulnerabilities
Hall 1In RussianComplexity -
I'll show you real-world examples of API attacks that can lead to system compromise, data disclosure, and unwanted actions. I will analyze several different attack cases: direct and indirect implementation of a prompt, role-playing, context change using contextual tokens.
Speakers
Invited experts
Alexandra Svatikova
T-Bank