AssurAI: Experience with Constructing Korean Socio-cultural Datasets to Discover Potential Risks of Generative AI

TL;DR

AssurAI is a comprehensive, quality-controlled Korean multimodal dataset designed to evaluate the safety of generative AI, addressing language and cultural gaps in existing safety datasets.

Contribution

We created AssurAI, a large-scale Korean multimodal safety dataset with a new risk taxonomy, rigorous quality control, and validation for assessing generative AI safety in Korean socio-cultural contexts.

Findings

AssurAI effectively assesses safety of recent LLMs.

The dataset covers 35 risk factors across multiple modalities.

Rigorous quality control ensures high data integrity.

Abstract

The rapid evolution of generative AI necessitates robust safety evaluations. However, current safety datasets are predominantly English-centric, failing to capture specific risks in non-English, socio-cultural contexts such as Korean, and are often limited to the text modality. To address this gap, we introduce AssurAI, a new quality-controlled Korean multimodal dataset for evaluating the safety of generative AI. First, we define a taxonomy of 35 distinct AI risk factors, adapted from established frameworks by a multidisciplinary expert group to cover both universal harms and relevance to the Korean socio-cultural context. Second, leveraging this taxonomy, we construct and release AssurAI, a large-scale Korean multimodal dataset comprising 11,480 instances across text, image, video, and audio. Third, we apply the rigorous quality control process used to ensure data integrity, featuring a two-phase construction (i.e., expert-led seeding and crowdsourced scaling), triple independent annotation, and an iterative expert red-teaming loop. Our pilot study validates AssurAI's effectiveness in assessing the safety of recent LLMs. We release AssurAI to the public to facilitate the development of safer and more reliable generative AI systems for the Korean community.