CaMeLs Can Use Computers Too: System-level Security for Computer Use Agents

TL;DR

This paper introduces Single-Shot Planning for Computer Use Agents, enabling secure, architecture-isolated task execution by precomputing complete plans, thus preventing prompt injection attacks while maintaining practical performance.

Contribution

It presents a novel planning approach that guarantees control flow integrity in CUAs, addressing the challenge of combining security with dynamic UI observation.

Findings

Precomputed execution graphs prevent instruction injections.

The approach retains up to 57% of frontier model performance.

Performance for smaller models improves by up to 19%.

Abstract

AI agents are vulnerable to prompt injection attacks, where malicious content hijacks agent behavior to steal credentials or cause financial loss. The only known robust defense is architectural isolation that strictly separates trusted task planning from untrusted environment observations. However, applying this design to Computer Use Agents (CUAs) -- systems that automate tasks by viewing screens and executing actions -- presents a fundamental challenge: current agents require continuous observation of UI state to determine each action, conflicting with the isolation required for security. We resolve this tension by demonstrating that UI workflows, while dynamic, are structurally predictable. We introduce Single-Shot Planning for CUAs, where a trusted planner generates a complete execution graph with conditional branches before any observation of potentially malicious content, providing provable control flow integrity guarantees against arbitrary instruction injections. Although this architectural isolation successfully prevents instruction injections, we show that additional measures are needed to prevent Branch Steering attacks, which manipulate UI elements to trigger unintended valid paths within the plan. We evaluate our design on OSWorld, and retain up to 57% of the performance of frontier models while improving performance for smaller open-source models by up to 19%, demonstrating that rigorous security and utility can coexist in CUAs.