Proofpoint Just Built Security for MCP. That Tells You Everything.

Proofpoint this week launched what they’re calling an Agent Integrity Framework — intent-based AI security that monitors whether an agent’s behavior aligns with the original request, defined policies, and intended purpose. It covers endpoints, browsers, and MCP agent connections specifically.

The product came from their acquisition of Acuvity.

Most AI security tooling so far has been reactive — block the bad prompt, flag the suspicious output, add guardrails around specific tools. Intent-based monitoring asks a different question: does what the agent is doing match what was asked? An agent that drifts from its task, takes actions outside its stated scope, or behaves differently across sessions — that’s the signal. And it’s the right threat model.

AI agents rarely do obviously malicious things. They do plausibly reasonable things that weren’t actually authorized. An agent that reads files it wasn’t asked to read, forwards context it shouldn’t have access to, or interprets ambiguous instructions in ways that benefit an attacker — none of that looks like a classic security event. It looks like normal agent behavior.

The MCP-specific coverage is what matters here. Proofpoint is a major enterprise vendor. They don’t build products for hypothetical attack surfaces. When they ship MCP security tooling, it means enterprise deployments are real enough to defend.

MCP has gone from developer toy to infrastructure — and real infrastructure attracts real security vendors, who don’t show up early.