Who controls your AI? (wrong question)

In a world of rapidly evolving agents, the technology “sovereignty” debate is leaving enterprises stuck in neutral.

Source: Author

This post is a summary of my recent in-depth article (free on Substack).

The technology sovereignty debate is not changing with the times. We are still focussed on infra (“who controls my compute”) and data (“who controls my data”), but in a world of agents, we need to think as carefully and deliberately about:

• Intelligence sovereignty: Who controls my models?
• Agency sovereignty: Who controls actions, or more specifically, who controls the agents that do work?

But enterprises struggle: the most frequent question I hear is “Should we use external AI” — this is the WRONG question (unless you are building your own models, you have no choice).

The RIGHT question is more nuanced: “Which parts of the intelligence and agency stack should I control”, and “which parts am I comfortable with external control.”

This is not an easy question though…

Take a look at my recent article on a deep dive on the emerging “sovereignty stack” and the questions it poses.

• More articles on agents on The Agentic Mesh Substack
• The Agentic Mesh Podcast channel on YouTube.
• The Agentic Mesh Podcast on Apple.
• The Agentic Mesh Podcast on Spotify.