I've spent 25 years building, breaking, and defending software systems. These days I think about what happens when autonomous AI agents need identity, access to secrets, and governance — and why the security frameworks we built for humans don't always transfer.
A practitioner's framework for securing agentic AI, told through a small-town metaphor. Start with Article 1 and read in order.
Article 1
Human identity frameworks don't map to agents. We need to start from scratch. The first article in the Building the Town series on securing agentic AI systems.
Article 2
If agent identity is assigned, not inherent, then the authority that assigns it is your most critical trust anchor. In a small town, that's the clerk's office.
Article 3
Access control and data sovereignty are fundamentally different security models. A locked office and a bank vault are not the same thing. Organizations need both.