智能体 AI 中的安全与隐私：重大挑战与未来方向
密码学与安全
2026-07-07 v1 人工智能
人机交互
摘要
我们基于一次地平线扫描活动，提出了智能体 AI 安全与隐私方面的关键挑战和未来研究方向。该活动汇集了来自学术界、工业界和政府的三十位领先国际专家，就与 AI 日益增长的自主性相关的新兴风险进行了重点讨论和协作练习。
引用
@article{arxiv.2607.06608,
title = {Security and Privacy in Agentic AI: Grand Challenges and Future Directions},
author = {Adam Jenkins and Agnieszka Kitkowska and Caterina Maidhof and Diego Paracuellos and Francesco Sovrano and Gonzalo Gabriel Mendez and Guillermo Suarez-Tangil and Hana Kopecka and Isabel Wagner and Isabel Barbera and Javier Carnerero-Cano and Jide Edu and Jose Luis Martin-Navarro and Jose Such and Josep Domingo-Ferrer and Juan Carlos Carrillo and Kopo Marvin Ramokapane and Mark Cote and Pablo Vellosillo and Ramon Ruiz-Dolz and Rongjun Ma and Ruba Abu-Salma and Sameer Patil and William Seymour and Xiao Zhan},
journal= {arXiv preprint arXiv:2607.06608},
year = {2026}
}