Who's Watching the Guardians?

With so much serious discussion happening in the world of AI, a little humour every now and then doesn’t hurt.

On October 19, Microsoft’s Hayete Gallot published an interesting article titled Securing the agentic insider.

The article makes an important point about the growing use of AI agents inside organizations. Just because we have deployed these agents ourselves doesn’t mean we should automatically trust everything they do.

As the article puts it:

“Agents are unpredictable.”

And then comes an interesting recommendation:

“To manage and control these insiders, we will need Guardian Agents”

These Guardian Agents would continuously monitor and supervise the actions of other AI agents, helping organizations detect risks before something goes wrong.

The security concern is quite real, especially as agents gain access to organizational data, tools and systems, and increasingly act on their own.

But I couldn’t help noticing the irony.

We started introducing AI agents to take work off our hands. Now we are introducing another set of agents to watch over the first set.

And if this continues, we might eventually end up creating an entire management hierarchy of AI agents, with agents supervising agents, who in turn supervise other agents.

All in the name of reducing the amount of supervision humans need to do.

Who’s watching the guardians?

Reference: Microsoft — Securing the agentic insider, October 19, 2026