1. Parser Differentials 2. I Built a RAG Chatbot. Then I Broke It

Why we picked it

*Talk Submissions*: [ruxmon@proton.me](http://ruxmon@proton.me%2A%2A/) **1\. Parser Differentials \- SiJing \(@moat\) and chsh** This is one of the bug classes we are always super excited to find and never tire from hearing about. In this talk we will go through a brief review of the latest literature around parser differentials, then examine a real case where an unchecked assumption could have caused huge problems in a complex cloud setup. **2\. I Built a RAG Chatbot\. Then I Broke It \- Sahil Devkota** Building an AI chatbot is exciting until someone tries to break it. In this talk, we'll explore prompt injection attacks in RAG systems, how attackers can manipulate AI to ignore its instructions, leak private data, or behave in unexpected ways. We'll also look at simple, practical ways to defend against these attacks. Whether you're a developer building AI applications or just curious about AI security, you'll leave with a practical understanding of how prompt injection works — an