Policy Compiler for Secure Agentic Systems
This addresses security and compliance challenges for organizations deploying AI agents in sensitive domains like customer service and healthcare, though it is an incremental improvement over existing policy enforcement methods.
The paper tackles the problem of enforcing complex authorization policies in LLM-based agentic systems by introducing PCAS, a policy compiler that models system state as a dependency graph and uses declarative rules for deterministic enforcement, improving policy compliance from 48% to 93% in customer service tasks with zero violations.
LLM-based agents are increasingly being deployed in contexts requiring complex authorization policies: customer service protocols, approval workflows, data access restrictions, and regulatory compliance. Embedding these policies in prompts provides no enforcement guarantees. We present PCAS, a Policy Compiler for Agentic Systems that provides deterministic policy enforcement. Enforcing such policies requires tracking information flow across agents, which linear message histories cannot capture. Instead, PCAS models the agentic system state as a dependency graph capturing causal relationships among events such as tool calls, tool results, and messages. Policies are expressed in a Datalog-derived language, as declarative rules that account for transitive information flow and cross-agent provenance. A reference monitor intercepts all actions and blocks violations before execution, providing deterministic enforcement independent of model reasoning. PCAS takes an existing agent implementation and a policy specification, and compiles them into an instrumented system that is policy-compliant by construction, with no security-specific restructuring required. We evaluate PCAS on three case studies: information flow policies for prompt injection defense, approval workflows in a multi-agent pharmacovigilance system, and organizational policies for customer service. On customer service tasks, PCAS improves policy compliance from 48% to 93% across frontier models, with zero policy violations in instrumented runs.