AI and Inference
Phase C A1a: split security/ + ai/ out of architecture/ (23 articles, 46 EN+ES files); category frontmatter fixed; redirects.yaml 301s; new _index.md landings
@@ -0,0 +1,25 @@ --- schema: foundry-doc-v1 title: "How AI Is Used and Contained" slug: ai-index category: ai type: topic content_type: topic quality: complete short_description: "Where AI sits and where it is not allowed: the boundary that keeps AI away from the authoritative record, the routing between models, and the small, customer-side models designed to learn a customer's own environment. The core runs fully without it." status: active bcsc_class: public-disclosure-safe last_edited: 2026-07-03 editor: pointsav-engineering paired_with: _index.es.md --- The **ai** category collects where AI sits in the platform and where it is not allowed. It covers the boundary that keeps AI away from the authoritative record, the routing between models, and the small, customer-side models designed to learn a customer's own environment. The deterministic core runs fully without AI. This is the front door for the platform's most distinctive architectural claim — AI is used, and it is contained — and for engineers looking up a specific piece of the AI stack: the inference boundary, sovereign routing, the vendor-tier model programme, and the training pipelines that produce it. ## See also - [Architecture](/architecture/) — the three-ring build that makes this boundary structural - [Building Blocks](/substrate/) — AI-adjacent mechanism concepts - [Platform Services](/services/) — the per-service pages, including the AI service itself