# Frontier AI Governance Map

> Interactive map of frontier AI governance covering six layers (international, national, sub-national, industry voluntary, corporate self-governance, infrastructure) and METR's nine common elements as an orthogonal filter. 6 June 2026 snapshot; 49 mechanisms across the stack with primary-source links, descriptions, and connections.

By Yuyu Shen
Personal site: https://buildwithwhy.com
Contact: buildwithwhy@gmail.com
Web version: https://ai-governance-map.buildwithwhy.com

## Documents

- [Full content as markdown](https://ai-governance-map.buildwithwhy.com/llms-full.txt) — every mechanism with description, context, primary source, coverage notes, and connections (49 mechanisms, 81 relationships)
- [Web version](https://ai-governance-map.buildwithwhy.com) — interactive HTML

## Conceptual structure

**Layers** organise by *who acts*:

- Layer 1 — International: Multilateral norms and summits
- Layer 2 — National regulation: Statutes, executive orders, agencies
- Layer 3 — Sub-national: US states acting where federal regulation is absent
- Layer 4 — Infrastructure: Regulators, AISIs, evaluators, compute controls
- Layer 5 — Industry voluntary: Cross-firm commitments and codes
- Layer 6 — Corporate self-governance: Frontier safety frameworks at each lab

**Coverage categories** (METR's nine common elements, applied as orthogonal filter):

- Capability thresholds: Levels at which obligations or risks trigger
- Model evaluations: Tests assessing model capabilities and risks
- Capability elicitation: Surfacing the full capabilities of a model
- Timing & frequency: When evaluations and obligations apply
- Weight security: Protecting model weights from theft
- Deployment mitigations: Refusals, monitoring, content controls
- Conditions for halting: When to pause development or deployment
- Accountability: Oversight, transparency, reporting
- Updating policies: How policies get revised over time