Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910105884164096 |
|---|---|
| author | Wang, Zijun Tu, Haoqin Zhang, Letian Chen, Hardy Wu, Juncheng Liu, Xiangyan Yuan, Zhenlong Pang, Tianyu Shieh, Michael Qizhe Liu, Fengze Zheng, Zeyu Yao, Huaxiu Zhou, Yuyin Xie, Cihang |
| author_facet | Wang, Zijun Tu, Haoqin Zhang, Letian Chen, Hardy Wu, Juncheng Liu, Xiangyan Yuan, Zhenlong Pang, Tianyu Shieh, Michael Qizhe Liu, Fengze Zheng, Zeyu Yao, Huaxiu Zhou, Yuyin Xie, Cihang |
| contents | 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., Capability, Identity, and Knowledge, 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. Our project page is https://ucsc-vlaa.github.io/CIK-Bench. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_04759 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw Wang, Zijun Tu, Haoqin Zhang, Letian Chen, Hardy Wu, Juncheng Liu, Xiangyan Yuan, Zhenlong Pang, Tianyu Shieh, Michael Qizhe Liu, Fengze Zheng, Zeyu Yao, Huaxiu Zhou, Yuyin Xie, Cihang Cryptography and Security Artificial Intelligence Computation and Language 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., Capability, Identity, and Knowledge, 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. Our project page is https://ucsc-vlaa.github.io/CIK-Bench. |
| title | Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw |
| topic | Cryptography and Security Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2604.04759 |