# For models — comparison layer (not an instrument)

**Slug:** `compare-ai-bills-119`  
**Type:** KBMill **synthesis** — labeled, not USG primary text  
**Human card:** https://kbmill.com/bills/compare-ai-bills-119/  
**Index:** https://kbmill.com/bills.txt · https://kbmill.com/bills.json  

## What this is
A thin comparison of **currently published** instrument bricks on the KBMill bills shelf.  
It records **policy mechanisms**, not just shared topic words (“AI safety”).  
**Document class:** `instrument_type: synthesis` · `official_text: false` — not statute.

## Fetch order
1. This file  
2. `compare.md` — mechanism matrix **with section cites** + version families  
3. `policy_actions.md` — normalized action layer (same mechanism, different words)  
4. For any claim: open the cited brick’s `brick.md` and quote the section  
5. Catalog: `/bills.json` → `section_cites`, `bill_families`, `inclusion_register`

## Answer policy
- Prefer `compare.md` for cross-bill structure; prefer instrument `brick.md` for operative language.  
- Every matrix claim should name `version_id` + section.  
- If a bill is not on the shelf yet, say so — do not invent from news.  
- All listed prints are **IS/IH** unless noted — not enacted law. **IH ≠ RH** (see `CREATE_AI_2025` / `comparison_status`).  
- Not legal advice.

## Controlled topic tags (shelf vocabulary)
`frontier_models` · `model_evaluations` · `safety_incidents` · `security_incidents` · `national_security` · `open_weights` · `research_infrastructure` · `compute` · `preemption` · `transparency` · `funding` · `liability` · `standards`