Is Safety Standard Same for Everyone? User-Specific Safety Evaluation of Large Language Models

2025-02-20

AI Summary

Current LLM safety evaluations rely on generic benchmarks, overlooking users’ personalized security requirements and thus failing at the individual level. To address this, we propose U-SAFEBENCH—the first user-specific safety benchmark—introducing and quantifying a user-profile-driven safety evaluation paradigm. It features an adversarial test suite spanning multi-dimensional privacy, ethical, and preference constraints. Empirical evaluation across 18 mainstream LLMs reveals widespread failure to satisfy user-level safety requirements. We further propose a lightweight Chain-of-Thought (CoT)-guided remediation method that explicitly models user profiles and constraints via safety-aware reasoning chains, improving average safety compliance by 23.6%. This work bridges the critical gap in personalized safety assessment and establishes a novel, scalable foundation for trustworthy LLM deployment and governance.

Technology Category

• Natural Language Processing: Safety and Robustness
• Philosophy and Ethics of AI: Safety, Robustness & Trustworthiness
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

As the use of large language model (LLM) agents continues to grow, their safety vulnerabilities have become increasingly evident. Extensive benchmarks evaluate various aspects of LLM safety by defining the safety relying heavily on general standards, overlooking user-specific standards. However, safety standards for LLM may vary based on a user-specific profiles rather than being universally consistent across all users. This raises a critical research question: Do LLM agents act safely when considering user-specific safety standards? Despite its importance for safe LLM use, no benchmark datasets currently exist to evaluate the user-specific safety of LLMs. To address this gap, we introduce U-SAFEBENCH, the first benchmark designed to assess user-specific aspect of LLM safety. Our evaluation of 18 widely used LLMs reveals current LLMs fail to act safely when considering user-specific safety standards, marking a new discovery in this field. To address this vulnerability, we propose a simple remedy based on chain-of-thought, demonstrating its effectiveness in improving user-specific safety. Our benchmark and code are available at https://github.com/yeonjun-in/U-SafeBench.

Problem

• Evaluate user-specific safety of LLMs
• Introduce U-SAFEBENCH for safety assessment
• Propose remedy using chain-of-thought technique

Innovation

• User-specific safety evaluation
• Chain-of-thought remedy
• U-SAFEBENCH benchmark