When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Life

TL;DR

This paper introduces SaLAD, a comprehensive multimodal safety benchmark for evaluating how well Large Language Models handle unsafe content in daily life, highlighting current limitations and proposing a safety-warning-based evaluation framework.

Contribution

The paper presents SaLAD, a new multimodal safety benchmark with real-world image-text samples, and proposes a safety-warning evaluation framework to better assess MLLMs' safety performance.

Findings

Top models achieve only 57.2% safe response rate on unsafe queries.

Current safety alignment methods have limited effectiveness.

Models struggle with identifying dangerous behaviors in daily life.

Abstract

As Multimodal Large Language Models (MLLMs) become an indispensable assistant in human life, the unsafe content generated by MLLMs poses a danger to human behavior, perpetually overhanging human society like a sword of Damocles. To investigate and evaluate the safety impact of MLLMs responses on human behavior in daily life, we introduce SaLAD, a multimodal safety benchmark which contains 2,013 real-world image-text samples across 10 common categories, with a balanced design covering both unsafe scenarios and cases of oversensitivity. It emphasizes realistic risk exposure, authentic visual inputs, and fine-grained cross-modal reasoning, ensuring that safety risks cannot be inferred from text alone. We further propose a safety-warning-based evaluation framework that encourages models to provide clear and informative safety warnings, rather than generic refusals. Results on 18 MLLMs demonstrate that the top-performing models achieve a safe response rate of only 57.2% on unsafe queries. Moreover, even popular safety alignment methods limit effectiveness of the models in our scenario, revealing the vulnerabilities of current MLLMs in identifying dangerous behaviors in daily life. Our dataset is available at https://github.com/xinyuelou/SaLAD.