Language models often answer harmful requests in low-resource languages (LRLs) that they refuse in high-resource languages (HRLs). Across three instruction-tuned models and 23 languages, harmful refusal falls from 87.9% in HRLs to 43.9% in LRLs, while harmless refusal remains low. A common explanation is that models represent harmfulness weakly in LRLs. We test whether harmfulness is instead represented but does not reliably produce refusal. Across three models, a harmfulness direction learned from HRL activations still separates harmful from harmless LRL prompts, showing that complete absence of harmfulness information cannot explain many failures. However, harmfulness scores shift downward for LRL prompts, making harmful prompts less likely to reach the range associated with refusal. Motivated by this shift, we train a low-rank logistic classifier on HRL activations and calibrate its threshold with a few target-language examples. During generation, the classifier conditionally adds or ablates the HRL harmfulness direction. With the same HRL data and 32 target-language examples per class, CAST remains limited by low harmful refusal and AdaSteer by high harmless refusal, yielding mean refusal selectivity ($Δ$ = harmful - harmless refusal) of 33.6 and 6.8, respectively. Our intervention reaches 54.5 while preserving MMLU utility. HRL-only calibration improves selectivity for Qwen and Gemma, whereas Llama benefits from target-language calibration. These results show that recalibrating existing representations can offer a training-free method for repairing low-resource safety failures.
LLM Safety Alignment in Low-Resource Languages: A Systematic Literature Review
African Institute for Mathematical Sciences (AIMS)(非洲数学科学研究所);Bayero University Kano(卡诺贝罗大学);Brown University(布朗大学);Centrum Wiskunde & Informatica(数学与计算机科学中心);University of Pretoria(比勒陀利亚大学);University of Hamburg(汉堡大学);Imperial College London(伦敦帝国学院)
When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models
Chinese Information Processing Laboratory, Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所信息处理实验室);University of Chinese Academy of Sciences(中国科学院大学)
SafeTutors: Benchmarking Pedagogical Safety in AI Tutoring Systems
Eindhoven University of Technology(埃因霍温理工大学);Indian Institute of Technology Kharagpur(印度理工学院Khargpur分校);Cisco Research(思科研究);Taras Shevchenko National University of Kyiv(塔拉斯·谢甫琴科基辅国立大学);National Technical University of Ukraine(乌克兰国家技术大学);New York University(纽约大学)
Manacá-1B-Instruct-2: Reproducible Safety Alignment for a Native Brazilian-Portuguese LLM. Negative DPO Results, a Controlled Composition Scan, and a Silent Failure Mode of the LLM Judge
National Laboratory for Scientific Computing (LNCC)(国家科学计算实验室);Institute of Artificial Intelligence(人工智能研究所)
Exploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Models
Mohamed Bin Zayed University of AI, UAE(穆罕默德·本·扎耶德人工智能大学,阿联酋);Khalifa University, UAE(卡比拉大学,阿联酋);Australian National University, Australia(澳大利亚国立大学,澳大利亚)
Uncertainty Is Not a Safety Net for Clinical VQA, but Can It Anticipate Model Failure?
Amsterdam University Medical Center, University of Amsterdam(阿姆斯特丹大学医学中心);Amsterdam Public Health(阿姆斯特丹公共卫生);LMU Munich(慕尼黑大学);Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心)