Step-level guards decide, before each tool call, whether an agent's next action is safe given the user's request and the trajectory so far. Users write those requests in many languages, while guard prompts and the benchmarks on which purpose-built guards are trained and evaluated are English. We ask whether a guard's verdict depends on the language of the request. We build LinguaGuardBench from the 1,220 labelled steps of TS-Bench AgentDojo-Traj: we translate only the user instruction into German, French, Turkish, Swahili and Amharic, and keep every tool call, observation and label byte-identical, so any change in a verdict is attributable to language alone. Six English rewordings and a whitespace placebo serve as controls, and managed guards are run three times to measure their noise. Across 19 guards, from frontier APIs to 4B open models and four purpose-built agent guards, most open, small and purpose-built guards change verdicts under translation beyond their own noise floor, and the effect grows with the rarity of the language: among the nine guards below 13B parameters, the mean placebo-adjusted flip rate rises from +1.5 pp on French to +4.4 pp on Amharic. Purpose-built guards are the most sensitive. The direction of the damage tends to follow guard size: open guards of 12-30B parameters block more benign steps (up to +5.0 pp false positives), and every significant loss of recall is in a guard below 10B (up to -7.4 pp). Frontier guards shift by at most about one point; the strongest, Gemini 3.8 Flash, returns identical verdicts with and without the request. The damage is concentrated in a few instructions and, in a post-hoc analysis, falls on steps where the agent's reasoning restates the English request, consistent with guards reacting to inconsistency between channels. English paraphrase moves every guard less than translation does. We release the benchmark and all 489k verdicts.
GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning
University of Georgia(佐治亚大学);University of Chicago(芝加哥大学);UIUC(伊利诺伊大学香槟分校);University of Texas at Austin(德克萨斯大学奥斯汀分校);University of California, Berkeley(加州大学伯克利分校);Emory University(埃默里大学);Virtue AI
EMBGuard: Constructing Hazard-Aware Guardrails for Safe Planning in Embodied Agents
Independent Researcher(独立研究者);Department of Biomedical Engineering(生物医学工程系);the Department of Intelligent Precision Healthcare Convergence, Sungkyunkwan University(智能精准医疗融合系,全州大学);Department of Artificial Intelligence, Yonsei University(人工智能系,延世大学)
JAILJUDGE: A Comprehensive Jailbreak Judge Benchmark with Multi-Agent Enhanced Explanation Evaluation Framework
The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));Baidu Inc.(百度公司);The Hong Kong University of Science and Technology(香港科技大学);University of Birmingham(伯明翰大学)
StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing
Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Beihang University(北京航空航天大学);Fudan University(复旦大学);Renmin University of China(中国人民大学);KAUST(阿卜杜拉国王科技大学)
GuardSec: A Multi-Modal Web Platform for Real-Time Digital Fraud Detection, Entity Verification, and Connection Security Analysis in the African Context
Université Marien Ngouabi(马里安·恩古瓦比大学);Société Informatique du Congo (SIC)(刚果计算机公司)
The Guard That Cried Wolf: How Scary Words Make Agent Guardrails Refuse Legitimate Actions
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络安全学院)
Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback
Southern University of Science and Technology(南方科技大学);University of Science and Technology of China(中国科学技术大学);University of Birmingham(伯明翰大学);Zhejiang University(浙江大学);East China Normal University(华东师范大学);Alibaba Group(阿里巴巴集团)