SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia

TL;DR

SEA-Guard introduces a culturally grounded multilingual safeguard framework tailored for Southeast Asia, addressing regional nuances in safety detection and outperforming existing models across multiple benchmarks.

Contribution

The paper presents a novel agentic data-generation framework and the SEA-Guard models, the first to incorporate SEA cultural contexts for multilingual safety in AI.

Findings

SEA-Guard outperforms existing safeguards in detecting regionally sensitive content.

The framework enables scalable creation of authentic, region-specific safety datasets.

SEA-Guard maintains strong general safety performance across benchmarks.

Abstract

Culturally aware safeguards are crucial for AI alignment in real-world settings, where safety extends beyond common sense and encompasses diverse local values, norms, and region-specific regulations. However, building large-scale, culturally grounded datasets is challenging due to limited resources and a scarcity of native annotators. Consequently, many safeguard models rely on machine translation of English datasets, often missing regional and cultural nuances. We present a novel agentic data-generation framework to scalably create authentic, region-specific safety datasets for Southeast Asia (SEA). On this foundation, we introduce the SEA-Guard family, the first multilingual safeguard models grounded in SEA cultural contexts. Evaluated across multiple benchmarks and cultural variants, SEA-Guard consistently outperforms existing safeguards at detecting regionally sensitive or harmful content while maintaining strong general safety performance.