AI agents are shipping to production faster than anyone can secure them. Prompt injection, tool abuse, system prompt leaks, multi-turn social engineering. The attack surface is wide open, and most defenders are flying blind.

On Friday, 5 June, Imperial College London and Pallma are putting that to the test.

You'll go head-to-head with a live AI financial assistant. Extracting secrets, hijacking tools, and chaining exploits to capture flags. No theory, no slides. Just you, the model, and a leaderboard.

What you'll walk away with

Hands-on experience attacking real LLM-powered agents
Exposure to the OWASP LLM Top 10, MITRE ATLAS, and OWASP Agentic Security in the wild
Challenges designed by the Pallma team, drawing on real vulnerabilities found in production AI agents
A spot on the scoreboard, and bragging rights

Who should come Imperial students, external researchers, and any cyber or AI professional who wants to understand how AI agents actually break. Whether you're a seasoned red teamer or curious about where AI security is heading, you'll learn something.

The essentials 🗓 Friday, 5 June 📍 Imperial College London 🍕 Food and drinks provided. Bring your laptop, leave hungry, we've got you.

Spots are limited. Lock yours in.

Location

Huxley Building

180 Queen's Gate, South Kensington, London SW7 2AZ, UK

Lecture Theatre 144

Hosted By

85 Going

AI