Secure Open-Source AI Agents with the DLP You Have Now

This proof-of-concept whitepaper describes a DLP-in-the-path architecture that integrates Symantec Distributed Detection Service (DDS) with open-source AI guardrails (e.g., Llama Guard) at an MCP gateway to inspect and enforce enterprise data protection policies on agent tool calls both inbound and outbound; it emphasizes fail-closed behavior, separation of detection and enforcement, practical inbound/outbound blocking examples, and the ability to reuse existing DLP policies to secure AI workflows rather than introducing a new, separate AI-specific data protection framework.