The-LLM-Red-Teamer-s-Playbook

The LLM Red Teamer's Playbook provides a systematic methodology for assessing and bypassing various defense layers in Large Language Models (LLMs), such as input filters and alignment mechanisms. It emphasizes a diagnostic approach to identify and understand the specific defenses in place before selecting appropriate attack techniques, mapped to the Adversarial AI Threat Modeling Framework (AATMF) v3. This guide is intended for AI red teamers, security engineers, and researchers, enabling them to conduct unauthorized testing responsibly while improving the security of AI systems.