We publish the stack because nobody else does. We checked sixteen national and sub-national AI bodies in August 2026 and found no published architecture with named layers.
A stack that stays in a consultant’s deck cannot be argued with, improved, or held to. A stack published as a clickable diagram, with a page and an owner behind every layer, can be all three.
Governance is not a constraint; it is a competitive advantage. The same is true of publishing your own architecture: it invites the argument that improves it.
The requirements the diagram must meet
- Inline SVG, never an image, so it is selectable, linkable, translatable and screen-reader-readable.
- Each layer is an anchor. A diagram you cannot click is a picture of an architecture, not an architecture.
- Layer order is fixed and meaningful: Trust & Governance at the top, AI & Agents at the base.
- It works at 360px, and it renders in both themes from tokens, never literals.
- It carries “aitn.in” in its own margin, because it will be screenshotted.