Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking?

2025/05/23 by Lu, Chengda, Fan, Xiaoyu, Huang, Yu +3 · 4 citations
#Artificial Intelligence (cs.AI) #FOS: Computer and information sciences

paper · doi:10.48550/arxiv.2505.17650

Abstract

Jailbreak attacks have been observed to largely fail against recent reasoning models enhanced by Chain-of-Thought (CoT) reasoning. However, the underlying mechanism remains underexplored, and relying solely on reasoning capacity may raise security concerns. In this paper, we try to answer the question: Does CoT reasoning really reduce harmfulness from jailbreaking? Through rigorous theoretical analysis, we demonstrate that CoT reasoning has dual effects on jailbreaking harmfulness. Based on the theoretical insights, we propose a novel jailbreak method, FicDetail, whose practical performance validates our theoretical findings.

Cited by

• Asynchronous Reasoning: Training-Free Interactive Thinking LLMs
• Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training
• A Comprehensive Survey on Trustworthiness in Reasoning with Large Language Models
• Argument-Based Consistency in Toxicity Explanations of LLMs

Related

• Chain-of-Thought Hijacking
• Hidden in Thought: Transferable Chain-of-Thought Artifacts Induce Harmful Behavior
• H-CoT: Hijacking the Chain-of-Thought Safety Reasoning Mechanism to Jailbreak Large Reasoning Models, Including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking
• Where Do Reasoning Models Refuse?
• Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training