Dark Prompts, Dirty Outputs: Jailbreaking LLMs for Red Team Ops
Val Saengphaibul
SAINTCON 2025 · Day 1 · Main Track 1
In this insightful SAINTCON talk, Val Saengphaibul, Director of Threat Response at Fortinet, navigates the rapidly evolving landscape of Large Language Models (LLMs) from a cybersecurity perspective. The presentation, titled "Dark Prompts, Dirty Outputs: Jailbreaking LLMs for Red Team Ops," delves into the critical need for security professionals to understand and leverage less-censored LLMs for effective red and purple teaming. Saengphaibul argues that while mainstream LLMs are heavily guarded by corporate policies and ethical frameworks, a new generation of open-source and locally hostable models offers unprecedented capabilities for simulating advanced cyber threats.