The Pitfalls of Over-Alignment: Overly Caution Health-Related Responses From LLMs are Unethical and Dangerous

ACL ARR 2026 January Submission6969 Authors

Keywords: Alignment, Over-Safety, LLM, Ethics

Abstract: Large Language Models (LLMs) are usually aligned with ``human values/preferences'' to prevent harmful output. However, in this paper, we argue that in health-related queries, over-alignment—leading to overly cautious responses—can itself be harmful, especially for people with anxiety and obsessive-compulsive disorder (OCD). This is not only unethical but also dangerous to the user, both mentally and physically. We also showed qualitative results that some LLMs exhibit varying degrees of alignment. Finally, we call for the development of LLMs that can provide more tailored and nuanced responses to health queries.

Paper Type: Short

Research Area: Ethics, Bias, and Fairness

Research Area Keywords: ethical considerations in NLP applications, reflections and critiques

Contribution Types: Position papers

Languages Studied: English

Submission Number: 6969