Embedded Files
Project Title: Multi Agent Security and Safety
Project Title: Multi Agent Security and Safety
We challenge existing AI/ML security and privacy practices. Our focus is on studying cutting-edge AI systems like assistants and agents. Can we trust them, what adversaries could do, what type of damage is possible? This is interdisciplinary work that touches on AI, LLMs, privacy and security.
Project Description:
We will design a multi-agent system that has access to various sensitive text data and implement attacks on the multi-agent systems to extract data or change systems goals.
Learning Objectives:
Critical thinking of different threat models, understanding of multi-agent systems, attacks and defenses in AI
Skills to learn:
NLP, software development
Report abuse