Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw

Zijun Wang, Haoqin Tu, Letian Zhang, Hardy Chen, Juncheng Wu, Xiangyan Liu, Zhenlong Yuan, Tianyu Pang, Michael Qizhe Shieh, Fengze Liu, Zeyu Zheng, Huaxiu Yao, Yuyin Zhou, Cihang Xie

Published: 23 May 2026, Last Modified: 23 May 2026ICML 2026 AIWILDEveryoneRevisionsBibTeXCC BY 4.0
Keywords: Agent, Safety, Real-World Evaluation
TL;DR: We introduce the CIK taxonomy (Capability, Identity, Knowledge) for personal AI agents' persistent state and conduct the first real-world evaluation showing that poisoning any dimension easily turns the agent from assistant to adversary.
Abstract: OpenClaw, the most widely deployed personal AI agent in early 2026, operates with full local system access and integrates with sensitive services such as Gmail, Stripe, and the filesystem. While these broad privileges enable high levels of automation and powerful personalization, they also expose a substantial attack surface that existing sandboxed evaluations fail to capture. To address this gap, we present the first real-world safety evaluation of OpenClaw and introduce the **CIK taxonomy**, which unifies an agent’s persistent state into three dimensions, i.e., **C**apability, **I**dentity, and **K**nowledge, for safety analysis. Our evaluations cover 12 attack scenarios on a live OpenClaw instance across four backbone models (Claude Sonnet 4.5, Opus 4.6, Gemini 3.1 Pro, and GPT-5.4). The results show that poisoning any single CIK dimension increases the average attack success rate from 24.6% to 64-74%, with even the most robust model exhibiting more than a threefold increase over its baseline vulnerability. We further assess three CIK-aligned defense strategies alongside a file-protection mechanism; however, the strongest defense still yields a 63.8% success rate under Capability-targeted attacks, while file protection blocks 97% of malicious injections but also prevents legitimate updates. Taken together, these findings show that the vulnerabilities are inherent to the agent architecture, necessitating more systematic safeguards to secure personal AI agents. We release the code at https://anonymous.4open.science/r/CIK-Bench-Anonymous-79E4.
Track: Regular Paper (9 pages)
Email Sharing: We authorize the sharing of all author emails with Program Chairs.
Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.
Submission Number: 81