SocialHarmBench: Revealing LLM Vulnerabilities to Socially Harmful Requests
This addresses safety gaps for LLM deployments in high-stakes sociopolitical settings, exposing systematic biases that could undermine human rights and democratic values, and is incremental as it builds on existing safety benchmarks.
The paper tackled the problem of LLM vulnerabilities in sociopolitical contexts by introducing SocialHarmBench, a dataset of 585 prompts across 7 categories and 34 countries, revealing that open-weight models like Mistral-7B show attack success rates up to 97-98% in areas such as historical revisionism and propaganda.
Large language models (LLMs) are increasingly deployed in contexts where their failures can have direct sociopolitical consequences. Yet, existing safety benchmarks rarely test vulnerabilities in domains such as political manipulation, propaganda and disinformation generation, or surveillance and information control. We introduce SocialHarmBench, a dataset of 585 prompts spanning 7 sociopolitical categories and 34 countries, designed to surface where LLMs most acutely fail in politically charged contexts. Our evaluations reveal several shortcomings: open-weight models exhibit high vulnerability to harmful compliance, with Mistral-7B reaching attack success rates as high as 97% to 98% in domains such as historical revisionism, propaganda, and political manipulation. Moreover, temporal and geographic analyses show that LLMs are most fragile when confronted with 21st-century or pre-20th-century contexts, and when responding to prompts tied to regions such as Latin America, the USA, and the UK. These findings demonstrate that current safeguards fail to generalize to high-stakes sociopolitical settings, exposing systematic biases and raising concerns about the reliability of LLMs in preserving human rights and democratic values. We share the SocialHarmBench benchmark at https://huggingface.co/datasets/psyonp/SocialHarmBench.