English
Related papers

Related papers: Auditing Agent Harness Safety

200 papers

When an LLM agent reads a confidential file, then writes a summary, then emails it externally, no single step is unsafe, but the sequence is a data leak. We call this safety drift: individually safe actions compounding into violations.…

Cryptography and Security · Computer Science 2026-03-31 Aditya Dhodapkar , Farhaan Pishori

Foundation models have transformed automated code generation, yet autonomous software-engineering agents remain unreliable in realistic development settings. The dominant explanation locates this gap in model capability. We propose a…

Software Engineering · Computer Science 2026-05-14 Hailin Zhong , Shengxin Zhu

As large language models (LLMs) evolve from static chatbots into autonomous agents, the primary vulnerability surface shifts from final outputs to intermediate execution traces. While safety guardrails are well-benchmarked for natural…

Cryptography and Security · Computer Science 2026-04-09 Yen-Shan Chen , Sian-Yao Huang , Cheng-Lin Yang , Yun-Nung Chen

Agents built on LLMs are increasingly deployed across diverse domains, automating complex decision-making and task execution. However, their autonomy introduces safety risks, including security vulnerabilities, legal violations, and…

Artificial Intelligence · Computer Science 2025-08-01 Haoyu Wang , Christopher M. Poskitt , Jun Sun

Third-party skills are becoming the package ecosystem for LLM agents. They package natural-language instructions, helper scripts, templates, documents, and service configuration into reusable workflows. This makes skills useful, but it also…

Cryptography and Security · Computer Science 2026-05-15 Haomin Zhuang , Hanwen Xing , Yujun Zhou , Yuchen Ma , Yue Huang , Yili Shen , Yufei Han , Xiangliang Zhang

The development of LLM-based autonomous agents for end-to-end software development represents a significant paradigm shift in software engineering. However, the scientific evaluation of these systems is hampered by significant challenges,…

Software Engineering · Computer Science 2025-11-07 Zhengran Zeng , Yixin Li , Rui Xie , Wei Ye , Shikun Zhang

To identify safety violations, auditors often search over large sets of agent traces. This search is difficult because failures are often rare, complex, and sometimes even adversarially hidden and only detectable when multiple traces are…

Artificial Intelligence · Computer Science 2026-04-14 Adam Stein , Davis Brown , Hamed Hassani , Mayur Naik , Eric Wong

AI agents powered by large language models (LLMs) are being deployed at scale, yet we lack a systematic understanding of how the choice of backbone LLM affects agent security. The non-deterministic sequential nature of AI agents complicates…

Cryptography and Security · Computer Science 2026-02-25 Julia Bazinska , Max Mathys , Francesco Casucci , Mateo Rojas-Carulla , Xander Davies , Alexandra Souly , Niklas Pfister

The integration of vision-language models (VLMs) is driving a new generation of embodied agents capable of operating in human-centered environments. However, as deployment expands, these systems face growing safety risks, particularly when…

Cryptography and Security · Computer Science 2025-10-21 Zonghao Ying , Le Wang , Yisong Xiao , Jiakai Wang , Yuqing Ma , Jinyang Guo , Zhenfei Yin , Mingchuan Zhang , Aishan Liu , Xianglong Liu

Unlike traditional automation tools or static LLM-based systems, agents combine decision-making and tool utilization to accomplish complex tasks, showing great potential in software engineering. However, existing studies largely focus on…

Software Engineering · Computer Science 2025-11-04 Zhuowen Yin , Cuifeng Gao , Chunsong Fan , Wenzhang Yang , Yinxing Xue , Lijun Zhang

EVMbench, released by OpenAI, Paradigm, and OtterSec, is the first large-scale benchmark for AI agents on smart contract security. Its results -- agents detect up to 45.6% of vulnerabilities and exploit 72.2% of a curated subset -- have…

Cryptography and Security · Computer Science 2026-03-12 Chaoyuan Peng , Lei Wu , Yajin Zhou

The performance of large language model (LLM) systems depends not only on model weights, but also on their harness: the code that determines what information to store, retrieve, and present to the model. Yet harnesses are still designed…

Artificial Intelligence · Computer Science 2026-03-31 Yoonho Lee , Roshen Nair , Qizheng Zhang , Kangwook Lee , Omar Khattab , Chelsea Finn

LLM-based agents are becoming increasingly capable, yet their safety lags behind. This creates a gap between what agents can do and should do. This gap widens as agents engage in multi-turn interactions and employ diverse tools, introducing…

Cryptography and Security · Computer Science 2026-02-17 Xu Li , Simon Yu , Minzhou Pan , Yiyou Sun , Bo Li , Dawn Song , Xue Lin , Weiyan Shi

Large Language Models (LLMs) are increasingly deployed as agents that operate in real-world environments, introducing safety risks beyond linguistic harm. Existing agent safety evaluations rely on risk-oriented tasks tailored to specific…

Artificial Intelligence · Computer Science 2026-02-04 Jingnan Zheng , Yanzhen Luo , Jingjun Xu , Bingnan Liu , Yuxin Chen , Chenhang Cui , Gelei Deng , Chaochao Lu , Xiang Wang , An Zhang , Tat-Seng Chua

As Large Language Model (LLM) agents become more widespread, associated misalignment risks increase. While prior research has studied agents' ability to produce harmful outputs or follow malicious instructions, it remains unclear how likely…

Modern LLM based agents are no longer passive text generators. They read repositories, call tools, browse the web, execute code, maintain memory, communicate with other agents, and act through long horizon workflows. This shift moves the…

Multiagent Systems · Computer Science 2026-05-12 Tianxiao Li , Yixing Ma , Haiquan Wen , Zhenglin Huang , Qianyu Zhou , Zeyu Fu , Guangliang Cheng

Autonomous agents powered by large language models (LLMs) show promising potential in assistive tasks across various domains, including mobile device control. As these agents interact directly with personal information and device settings,…

Machine Learning · Computer Science 2026-01-28 Juyong Lee , Dongyoon Hahm , June Suk Choi , W. Bradley Knox , Kimin Lee

Recent advances in Large Language Models (LLMs) have sparked concerns over their potential to acquire and misuse dangerous or high-risk capabilities, posing frontier risks. Current safety evaluations primarily test for what a model…

Computers and Society · Computer Science 2025-11-27 Udari Madhushani Sehwag , Shayan Shabihi , Alex McAvoy , Vikash Sehwag , Yuancheng Xu , Dalton Towers , Furong Huang

The rapid proliferation of large language models (LLMs) has intensified the requirement for reliable safety evaluation to uncover model vulnerabilities. To this end, numerous LLM safety evaluation benchmarks are proposed. However, existing…

Computation and Language · Computer Science 2025-08-22 Xiangyang Zhu , Yuan Tian , Chunyi Li , Kaiwei Zhang , Wei Sun , Guangtao Zhai

Agent benchmarks have become the de facto measure of frontier AI competence, guiding model selection, investment, and deployment. However, reward hacking, where agents maximize a score without performing the intended task, emerges…

Artificial Intelligence · Computer Science 2026-05-14 Hao Wang , Hanchen Li , Qiuyang Mang , Alvin Cheung , Koushik Sen , Dawn Song