Workshop on Hacking LLM Guardrails

In this workshop, participants will get hands-on experience both building and breaking AI guardrails.

Working in teams, you’ll design guardrail policies to defend sensitive data within a vulnerable AI application, while attempting to bypass those of others.

Who will win? The scoreboard will tell, but in the end everyone walks away with a deeper grasp of how LLMs can be manipulated and the challenges of building robust guardrails.

Requirements:

• Laptop that can SSH into external machine.
• Local Python environment and an IDE (both optional).

Time: November 26, from 13:30 to 17:00

Venue: Sirris' Leuven office - Address: Gaston Geenslaan 8, 3001 Leuven

Organization: This workshop is organized by KU Leuven and Sirris, with the support of Vlaio and the Cybersecurity Program Flanders.