Sovereign AI: what should a company own when it runs agents?

[jenna-winkler]

We just published The Seven Pillars of Sovereign Enterprise AI on the AI Industrialization blog. We'd like to use this thread to hear how these ideas resonate with people building and running agents.

The idea

Sovereign AI is about having control over how AI operates within your organization, and the flexibility to change things as your needs evolve.

That means being able to decide where your data goes, what rules your agents follow, what they cost, and which models and tools they use. It also means being able to switch models, tools, or vendors without having to rebuild everything from scratch.

The post breaks this down into seven pillars: boundaries, security policy, records, costs, models and harnesses, building blocks, and knowledge you can take with you.

What we'd like to hear

• Which of the seven matters most to your team right now? Which feels less urgent?
• Where have you run into lock-in, whether with a model, harness, vendor, or tool you can't easily move away from?
• If you moved an agent to another platform, what would you need to bring with you?
• What's missing from the list?

We'd love to hear where this framework resonates, where it falls short, and what you've learned from running agents in your own environment.