Elamur Research⌕
☰
home / papers / 2026 / SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia

SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia

Paper · 2026 · Panuthep Tasawong, Jian Gang Ngui, Alham Fikri Aji et al.

Summary

Most safety classifiers that screen LLM inputs and outputs are built for English. Multilingual ones usually depend on machine-translated English data, which serves Southeast Asian languages poorly and misses local cultural, religious and regulatory sensitivities. The authors address this gap with an agentic synthesis pipeline:

  • A guideline agent turns metadata (cultural topic, country, prompt type, label type) into detailed annotation-style instructions.
  • A generator combines these instructions with a persona and a target language to write culturally specific prompts, which are paraphrased to reduce keyword bias.
  • Several LLMs then write responses.
  • Labels and quality filtering come from a Monte Carlo Reasoning Ensemble (MCRE). MCRE runs repeated stochastic reasoning passes and turns the vote frequencies into soft labels.

The resulting data trains the SEA-Guard family of 4B, 8B and 12B models. SEA-Guard reaches state-of-the-art results on a SEA cultural safety benchmark and stays competitive on generic multilingual safety, even without generic training data. It also transfers zero-shot to vision-text safety tasks. This offers a scalable route to culturally aware safeguards for low-resource regions.

Key points

  • Coverage and data size: the data spans 8 languages, namely Vietnamese, Thai, Tamil, Tagalog, Malay, Indonesian, English and Burmese, with about 870K synthetic samples per language across 53 SEA cultural categories.
  • Generation models: prompts come from Gemma-SEA-LION-v4-27B. Responses come from Llama3.1-70B-IT, Gemma3-27B-IT, Gemma-SEA-LION-v4-27B-IT and GPT-OSS-20B-IT.
  • Labeling with MCRE: N = 10 passes vote over a five-level ordinal scale, which is collapsed into Safe / Sensitive / Harmful labels. Native-speaker reviewers (32 graduate students) accepted 79.51% of sampled labels.
  • Zero-shot transfer: SEA-Guard beats the baseline in 6 of 7 vision-language safety benchmarks. It also proves robust to adversarial attacks and to both over- and under-blocking.
  • Release and limitations: models and data are released under MIT on Hugging Face (aisingapore/sea-guard). Khmer and Lao are not covered because no benchmarks exist for them.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢GoogleMountain View · United States430 works · 424,363 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksSocial Sciences · 1,772 worksSafety Research · 543 works

🧭 Topics 2Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 worksBias, fairness & safety · 517 works

Cite

Cite as Panuthep Tasawong, Jian Gang Ngui, Alham Fikri Aji, Trevor Cohn and Peerat Limkonchotiwat. 2026. SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.141

@inproceedings{tasawong2026sea,
  title = {{SEA-Guard: Culturally Grounded Multilingual Safeguard for Southeast Asia}},
  author = {Panuthep Tasawong, Jian Gang Ngui, Alham Fikri Aji, Trevor Cohn and Peerat Limkonchotiwat},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.141},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.141}
}

References

45 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Azure. 2025. Azure ai content safety documentation. find
  2. Samuel Cahyawijaya, Holy Lovenia, Joel Ruben Antony Moniz, Tack Hwa Wong, Mohammad Rifqi Farhansyah, Thant Thiri Maung, Frederikus Hudi, David Anugraha, Muhammad Ravi Shulthan Habibi, Muhammad Reza Qorib, Amit Agarwal, Joseph Marvin Imperial, Hitesh Laxmichand Patel, Vicky Feliren, Bahrul Ilmi Nasut… find
  3. Hakan Inan, Kartikeya Upasani, Jianfeng Chi, Rashi Rungta, Krithika Iyer, Yuning Mao, Michael Tontchev, Qing Hu, Brian Fuller, Davide Testuggine, and Madian Khabsa. 2023. Llama guard: Llm-based input-output safeguard for human-ai conversations. Preprint, arXiv:2312.06674.
  4. Patrick Chao, Edoardo Debenedetti, Alexander Robey, Maksym Andriushchenko, Francesco Croce, Vikash Sehwag, Edgar Dobriban, Nicolas Flammarion, George J. Pappas, Florian Tramer, Hamed Hassani, and Eric Wong. 2024a. Jailbreakbench: An open robustness benchmark for jailbreaking large language models. P…
  5. Liwei Jiang, Kavel Rao, Seungju Han, Allyson Ettinger, Faeze Brahman, Sachin Kumar, Niloofar Mireshghallah, Ximing Lu, Maarten Sap, Yejin Choi, and Nouha Dziri. 2024. Wildteaming at scale: From in-the-wild jailbreaks to (adversarially) safer language models. Preprint, arXiv:2406.18510.
  6. Patrick Chao, Alexander Robey, Edgar Dobriban, Hamed Hassani, George J. Pappas, and Eric Wong. 2024b. Jailbreaking black box large language models in twenty queries. Preprint, arXiv:2310.08419.
  7. Yihe Deng, Yu Yang, Junkai Zhang, Wei Wang, and Bo Li. 2025. Duoguard: A two-player rl-driven framework for multilingual llm guardrails. Preprint, arXiv:2502.05163. — [arXiv:2502.05163](https://arxiv.org/abs/2502.05163) find
  8. Stefan Evert. 2004. The statistics of word cooccurrences: Word pairs and collocations. find
  9. Jiaming Ji, Donghai Hong, Borong Zhang, Boyuan Chen, Juntao Dai, Boren Zheng, Tianyi Qiu, Jiayi Zhou, Kaile Wang, Boxuan Li, Sirui Han, Yike Guo, and Yaodong Yang. 2025. Pku-saferlhf: Towards multi-level safety alignment for llms with human preference. Preprint, arXiv:2406.15513.
  10. Raviraj Joshi, Rakesh Paul, Kanishk Singla, Anusha Kamath, Michael Evans, Katherine Luna, Shaona Ghosh, Utkarsh Vaidya, Eileen Long, Sanjay Singh Chauhan, and Niranjan Wartikar. 2025. Cultureguard: Towards culturally-aware dataset and guard model for multilingual safety applications. Preprint, arXiv… find
  11. Priyanshu Kumar, Devansh Jain, Akhila Yerukola, Liwei Jiang, Himanshu Beniwal, Thomas Hartvigsen, and Maarten Sap. 2025. Polyguard: A multilingual safety moderation tool for 17 languages. In Second Conference on Language Modeling. find
  12. Ruixiang Feng, Shen Gao, Xiuying Chen, Lisi Chen, and Shuo Shang. 2025. CulFiT: A fine-grained cultural-aware LLM training paradigm via multilingual critique data synthesis. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2241… find
  13. LakeraAI. 2025. Lakeraguard. find
  14. Jiahui Geng, Qing Li, Zongxiong Chen, Yuxia Wang, Derui Zhu, Zhuohan Xie, Chenyang Lyu, Xiuying Chen, Preslav Nakov, and Fakhri Karray. 2025. VSCBench: Bridging the gap in vision-language model safety calibration. In Findings of the Association for Computational Linguistics: ACL 2025, pages 3047–305… find
  15. Yue Liu, Hongcheng Gao, Shengfang Zhai, Jun Xia, Tianyi Wu, Zhiwei Xue, Yulin Chen, Kenji Kawaguchi, Jiaheng Zhang, and Bryan Hooi. 2025. Guardreasoner: Towards reasoning-based LLM safeguards. In ICLR 2025 Workshop on Foundation Models in the Wild. find
  16. Google Google Cloud. 2025. Model armor overview. find
  17. Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, and Nouha Dziri. 2024. Wildguard: Open one-stop moderation tools for safety risks, jailbreaks, and refusals of llms. Preprint, arXiv:2406.18495.
  18. John Hughes, Sara Price, Aengus Lynch, Rylan Schaeffer, Fazl Barez, Sanmi Koyejo, Henry Sleight, Erik Jones, Ethan Perez, and Mrinank Sharma. 2024. Best-of-n jailbreaking. Preprint, arXiv:2412.03556. — [arXiv:2412.03556](https://arxiv.org/abs/2412.03556) find
  19. Cheng Li, Mengzhuo Chen, Jindong Wang, Sunayana Sitaram, and Xing Xie. 2024. Culturellm: Incorporating cultural differences into large language models. In Advances in Neural Information Processing Systems, volume 37, pages 84799–84838. Curran Associates, Inc.
  20. Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar, Lester James V. Miranda, Jennifer Santoso, Elyanah Aco, Akhdan Fadhilah, Jonibek Mansurov, Joseph Marvin Imperial, Onno P. Kampman, Joel Ruben Antony Moniz, Muhammad Ravi Shulthan Habibi, Frederikus Hudi, Railey Montalan, Ryan Ignatius, Joanito… — SEACrowd: A Multilingual Multimodal Data Hub and Benchmark Suite for Southeast Asian Languages
  21. Raymond Ng, Thanh Ngan Nguyen, Yuli Huang, Ngee Chia Tai, Wai Yi Leong, Wei Qi Leong, Xianbin Yong, Jian Gang Ngui, Yosephine Susanto, Nicholas Cheng, Hamsawardhini Rengarajan, Peerat Limkonchotiwat, Adithya Venkatadri Hulagadri, Kok Wai Teng, Yeo Yeow Tong, Bryan Siow, Wei Yi Teo, Wayne Lau, Choon… — SEA-LION: Southeast Asian Languages in One Network
  22. Xuan-Phi Nguyen, Wenxuan Zhang, Xin Li, Mahani Aljunied, Zhiqiang Hu, Chenhui Shen, Yew Ken Chia, Xingxuan Li, Jianyu Wang, Qingyu Tan, Liying Cheng, Guanzheng Chen, Yue Deng, Sen Yang, Chaoqun Liu, Hang Zhang, and Lidong Bing. 2024. SeaLLMs - large language models for Southeast Asia. In Proceedings…
  23. Jean De Dieu Nyandwi, Yueqi Song, Simran Khanuja, and Graham Neubig. 2025. Grounding multilingual multimodal LLMs with cultural knowledge. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 24198– 24242, Suzhou, China. Association for Computational Lingu…
  24. OpenAI. 2024. Upgrading the moderation api with our new multi- modal moderation model. find
  25. Yuqi Ren and Deyi Xiong. 2023. HuaSLIM: Human attention motivated shortcut learning identification and mitigation for large language models. In Findings of the Association for Computational Linguistics: ACL 2023, pages 12350–12365, Toronto, Canada. Association for Computational Linguistics. find
  26. Wenliang Shan, Michael Fu, Rui Yang, and Chakkrit Tantithamthavorn. 2025. Sealguard: Safeguarding the multilingual conversations in southeast asian languages for llm software systems. Preprint, arXiv:2507.08898. — [arXiv:2507.08898](https://arxiv.org/abs/2507.08898) find
  27. Yosephine Susanto, Adithya Venkatadri Hulagadri, Jann Railey Montalan, Jian Gang Ngui, Xianbin Yong, Wei Qi Leong, Hamsawardhini Rengarajan, Peerat Limkonchotiwat, Yifan Mai, and William Chandra Tjhi. 2025. SEA-HELM: Southeast Asian holistic evaluation of language models. In Findings of the Associat… find
  28. Leanne Tan, Gabriel Chua, Ziyu Ge, and Roy Ka-Wei Lee. 2025. LionGuard 2: Building lightweight, data-efficient & localised multilingual content moderators. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: System Demonstrations, pages 264–285, Suzhou, China.… find
  29. Panuthep Tasawong, Napat Laosaengpha, Wuttikorn Ponwitayarat, Sitiporn Lim, Potsawee Manakul, Samuel Cahyawijaya, Can Udomcharoenchaikit, Peerat Limkonchotiwat, Ekapol Chuangsuwanich, and Sarana Nutanong. 2025a. Shortcut learning in safety: The impact of keyword bias in safeguards. In Proceedings of… find
  30. Panuthep Tasawong, Jian Gang Ngui, Alham Fikri Aji, Trevor Cohn, and Peerat Limkonchotiwat. 2025b. Sea-safeguardbench: Evaluating ai safety in sea languages and cultures. Preprint, arXiv:2512.05501. — [arXiv:2512.05501](https://arxiv.org/abs/2512.05501) find
  31. Nandan Thakur, Jianmo Ni, Gustavo Hernandez Abrego, John Wieting, Jimmy Lin, and Daniel Cer. 2024. Leveraging LLMs for synthesizing training data across many languages in multilingual dense retrieval. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computat… find
  32. Bibek Upadhayay and Vahid Behzadan. 2025. X-guard: Multilingual guard agent for content moderation. In Proceedings of the The First Workshop on LLM Security (LLMSEC), pages 54–86, Vienna, Austria. Association for Computational Linguistics. find
  33. Bibek Upadhayay, Vahid Behzadan, and Ph. D. 2025. X-guard: Multilingual guard agent for content moderation. Preprint, arXiv:2504.08848. find
  34. Sahil Verma, Keegan Hines, Jeff Bilmes, Charlotte Siska, Luke Zettlemoyer, Hila Gonen, and Chandan Singh. 2025. MULTIGUARD: An efficient approach for AI safety moderation across languages and modalities. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages… find
  35. Tianlu Wang, Rohit Sridhar, Diyi Yang, and Xuezhi Wang. 2022. Identifying and mitigating spurious correlations for improving robustness in NLP models. In Findings of the Association for Computational Linguistics: NAACL 2022, pages 1719–1729, Seattle, United States. Association for Computational Ling… find
  36. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-thought prompting elicits reasoning in large language models. In Proceedings of the 36th International Conference on Neural Information Processing Systems, NIPS ’22, R…
  37. Yahan Yang, Soham Dan, Shuo Li, Dan Roth, and Insup Lee. 2025. MrGuard: A multilingual reasoning guardrail for universal LLM safety. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 27365–27384, Suzhou, China. Association for Computational Linguistics. find
  38. Yahan Yang, Soham Dan, Dan Roth, and Insup Lee. 2024. Benchmarking llm guardrails in handling multilingual toxicity. Preprint, arXiv:2410.22153. find
  39. Wenqian Ye, Luyang Jiang, Eric Xie, Guangtao Zheng, Yunsheng Ma, Xu Cao, Dongliang Guo, Daiqing Qi, Zeyu He, Yijun Tian, Megan Coffee, Zhe Zeng, Sheng Li, Ting-hao, Huang, Ziran Wang, James M. Rehg, Henry Kautz, and Aidong Zhang. 2025. The clever hans mirage: A comprehensive survey on spurious corre… — [arXiv:2402.12715](https://arxiv.org/abs/2402.12715) find
  40. Xiang Yue, Yueqi Song, Akari Asai, Seungone Kim, Jean de Dieu Nyandwi, Simran Khanuja, Anjali Kantharuban, Lintang Sutawika, Sathyanarayanan Ramamoorthy, and Graham Neubig. 2025. Pangea: A fully open multilingual multimodal LLM for 39 languages. In The Thirteenth International Conference on Learning… find
  41. Wenjun Zeng, Yuchi Liu, Ryan Mullins, Ludovic Peran, Joe Fernandez, Hamza Harkous, Karthik Narasimhan, Drew Proud, Piyush Kumar, Bhaktipriya Radharapu, Olivia Sturman, and Oscar Wahltinez. 2024. Shieldgemma: Generative ai content moderation based on gemma. Preprint, arXiv:2407.21772. — [arXiv:2407.21772](https://arxiv.org/abs/2407.21772) find
  42. Xinyu Zhang, Pei Zhang, Shuang Luo, Jialong Tang, Yu Wan, Baosong Yang, and Fei Huang. 2025. Cul-tureSynth: A hierarchical taxonomy-guided and retrieval-augmented framework for cultural question-answer synthesis. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 10448–1… find
  43. Haiquan Zhao, Chenhan Yuan, Fei Huang, Xiaomeng Hu, Yichang Zhang, An Yang, Bowen Yu, Dayiheng Liu, Jingren Zhou, Junyang Lin, Baosong Yang, Chen Cheng, Jialong Tang, Jiandong Jiang, Jianwei Zhang, Jijie Xu, Ming Yan, Minmin Sun, Pei Zhang, and 24 others. 2025. Qwen3guard technical report. Preprint,…
  44. Kaiwen Zhou, Chengzhi Liu, Xuandong Zhao, Anderson Compalas, Dawn Song, and Xin Eric Wang. 2025. Multimodal situational safety. In The Thirteenth International Conference on Learning Representations. find
  45. Yongshuo Zong, Ondrej Bohdal, Tingyang Yu, Yongxin Yang, and Timothy Hospedales. 2024. Safety fine-tuning at (almost) no cost: a baseline for vision large language models. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tasawong-2026-sea-guard-culturally-grounded-multilingual/