Mandiant opens its agentic security harness to the wider industry

Google’s Mandiant is making its internal framework for running and auditing security agents available to organisations outside its own team. Put simply, it’s a structure for managing AI agents that carry out security work, and for checking what those agents did, how they behaved, and whether they stayed within the rules set for them.

That matters because security teams are starting to use AI for defensive tasks, not just experimenting with chatbots or using AI to write code. An agent that can investigate alerts, gather evidence, or support incident response could save time. But it also needs clear limits, records of its actions, and a way for people to review its decisions. Without that, companies risk deploying systems they can’t properly supervise, especially in an area as sensitive as security.

Mandiant’s move sits within a broader push to establish more consistent ways to deploy defensive AI agents. For business owners, the useful signal is not that you need security agents tomorrow. It’s that the practical foundations are being built now: governance, audit trails, permissions, and human oversight. Those are the things to ask about when a vendor says their AI can “handle” security work.

This is the kind of thing we build into an AI command centre, where AI activity should be visible, governed, and tied to accountable business processes.