Architecting Secure AI Agents: Perspectives on System-Level Defenses Against Indirect Prompt Injection Attacks

A position paper proposes system-level defenses against indirect prompt injection attacks in AI agents powered by large language models, arguing that dynamic replanning and security policy updates are often necessary for dynamic tasks and realistic environments. The paper's authors outline three positions, including that context-dependent security decisions should only be made within system designs that strictly constrain what a model can observe and decide, and that personalization and human interaction should be core design considerations in ambiguous cases. The paper also discusses limitations of existing benchmarks that can create a false sense of utility and security.

AI agents, predominantly powered by large language models LLMs , are vulnerable to indirect prompt injection, in which malicious instructions embedded in untrusted data can trigger dangerous agent actions. This position paper discusses our vision for system-level defenses against indirect prompt injection attacks. We articulate three positions: 1 dynamic replanning and security policy updates are often necessary for dynamic tasks and realistic environments; 2 certain context-dependent security decisions would still require LLMs or other learned models , but should only be made within system designs that strictly constrain what the model can observe and decide; 3 in inherently ambiguous cases, personalization and human interaction should be treated as core design considerations. In addition to our main positions, we discuss limitations of existing benchmarks that can create a false sense of utility and security. We also highlight the value of system-level defenses, which serve as the skeleton of agentic systems by structuring and controlling agent behaviors, integrating rule-based and model-based security checks, and enabling more targeted research on model robustness and human interaction.