This session delves into AI red teaming with a focus on identifying and mitigating vulnerabilities in large language models (LLMs). Participants will engage with an intentionally vulnerable LLM, exploring real- world adversarial scenarios where AI systems can be exploited. Using cutting-edge open-source and proprietary tools such as Protect AI’s Model Scan, Rebuff, LLMGuard, NeMo Guardrails, and Garak, i'll will demonstrate how to detect and exploit vulnerabilities in LLMs. Attendees will gain practical experience with these tools, learning how to assess model security, apply guardrails, and defend against adversarial attacks.
By the end of this session, participants will:
Understand how to perform AI red teaming on LLMs.
Use Model Scan, Rebuff, LLMGuard, NeMo Guardrails, and Garak for detecting and mitigating vulnerabilities.
Gain hands-on skills to apply these techniques in their own environments, improving the robustness and security of AI systems.
