Google publishes workshop report on agent privacy and security
What happened
On 6 October 2026, Google Research published the report "Open and Emerging Questions in Agent Privacy and Security: A Contextual Perspective". The report was produced by more than 50 people from academia and industry at the CAPS workshop held in New York in late 2025. It argues that agents differ from traditional software in three ways: unstructured interfaces, probabilistic control flow and autonomous delegation. It advocates building a contextual policy engine that generates policies in real time based on the theory of Contextual Integrity, with coordinated defences across system sandboxes, model inference and user controls.
Written by AI from the reports · updated 10 d ago
Timeline
Follow the reports to see the event from each side.
- Google ResearchOpen and Emergent Problems in Agentic Privacy and Security: A Contextual Angle
We're rapidly transitioning to a computing landscape defined by highly general, increasingly autonomous agents. Driven by large language models (LLMs) that can dynamically generate plans and invoke external tools, these systems offer the potential for AI to seamlessly handle complex, multi-step tasks on our behalf.