Large Language Models (LLMs) are susceptible to sophisticated multi-turn jailbreak attacks that exploit their reasoning capabilities. The attack, known as Reasoning-Augmented Conversation Exploit (RACE), transforms harmful queries into seemingly benign reasoning tasks. By leveraging the LLM's advanced reasoning abilities, attackers can ultimately induce the model to generate unsafe content. This method effectively circumvents standard safety mechanisms designed to prevent the creation of harmful responses. Mitigation steps: **For AI Developers:** * [Implement a layered security approach with multiple safety checks at different stages of the query processing pipeline] * [Strengthen safety mechanisms beyond simple keyword filtering to detect and prevent reasoning-based attacks] **For Model Trainers/Fine-tuners:** * [Develop robust models that can identify and resist manipulation of their reasoning processes] * [Develop more sophisticated detection methods to identify reasoning-based attacks, including analysis of information gain during conversation] * [Conduct rigorous red-teaming and adversarial testing to identify and address vulnerabilities before deployment]