Google Research outlines open problems in agentic privacy and security
A Google Research blog report proposes a multi-layered approach to agentic privacy and security spanning system, model, user and ecosystem levels. At the system level it explores dynamically limiting agent capabilities inside sandboxes, establishing agent identity, and granting or revoking data access as context changes; at the model level it argues agents handling complex tasks must judge whether actions such as sharing user data are appropriate and disambiguate under-specified prompts. The report also calls for dynamic, contextual and personalized controls to replace the notice-and-choice framework's assumption of foreseeability, guardrails against collusion in multi-agent interactions, and governance work on sharing contextual norms, resolving conflicts and verifying compliance.
The report argues autonomous agents' behavior is inherently probabilistic, high volume and generative, while the traditional notice-and-choice framework relies on an assumption of foreseeability, motivating a shift toward dynamic, contextual user controls.
Multi-layered approach to agentic privacy and security
Working towards contextual policy engines together with necessary advances at the model- and user-level results in a multi-layered approach. Our report outlines opportunities for innovation across the entire stack:
- System-level sandboxing: The execution environment can impose additional guardrails to limit the impact of agent or model failures and facilitate detection and response. Traditional operating systems sandbox applications by statically limiting access, say to files or networks. For agents, we explore dynamically limiting agentic capabilities, establishing agent identity, and authorizing or revoking access to data based on changes in context.
- Model-level reasoning: An agent performing complex tasks has to judge the appropriateness of its actions, such as sharing user data. Future research should enable models to disambiguate under-specified user prompts and reason about appropriateness under changing contextual norms.
- User-centric controls: The traditional “Notice and Choice” framework assumes that individuals can make fine-grained decisions about actions affecting personal data and relies on an assumption of foreseeability. However, autonomous agents’ behavior is inherently probabilistic, high volume, and generative. We explore shifting interfaces towards dynamic, contextual, and personalized control mechanisms consistent with user mental models.
- Multi-agent interactions: So that AI agents can safely collaborate with each other on complex tasks we need to develop effective guardrails to prevent the agents from colluding and violating contextual norms.
- Ecosystem governance: We explore challenges in governance mechanisms to arrive at a set of contextual norms that are collected and shared across agents. Furthermore, as norms can differ across domains and entities we need mechanisms that allow us to resolve conflicts, negotiate norm changes, and verify compliance.