As AI agents gain prevalence, users are increasingly exposed to the risks such systems entail. Prompt injection attacks and hallucinations can cause agents to leak private information to third parties. Additionally, as autonomous systems, agents pose the risk of performing sensitive tasks, such as bank transactions, without user intent or authorization. To address this challenge, the agentic security community has proposed numerous secure agentic systems. Much of this work has focused on product-level approaches, where agentic system developers apply the same security policies and permissions to all users. However, different users have varying needs and preferences, necessitating support for user-level permissions policies in agentic AI systems.
To understand how user-level permissions are handled in AI agent systems, we surveyed 21 proposals for agent permissions systems. From this review, we constructed a taxonomy of how different systems specify user-level permissions policies, both at the user interface and internally; derive internal policies from user input; and enforce those policies at run-time. We then analyzed five prominent commercial agents and compared their permissions handling to agentic permissions systems in the literature. We identified several high-level themes across the literature and commercial agents, as well as multiple gaps where future work is needed.
Blogger's Review: This article delves into the challenges of user permission management in AI agents, emphasizing the importance of customization. As AI technology advances, balancing user privacy with agent autonomy will be a critical issue for future development. This research provides valuable insights for improving agent security.