The fragmented AI governance literature, mapped.
9 major AI-governance instruments × 12 contested topics. Every cell traceable to its primary source. Updated as instruments evolve.
| Topic ↓ / Instrument → | EU-AIA | US-EO | UK-WHITEPAPER | CN-GENAI | G7-HIROSHIMA | OECD-AI | COE-AI | UN-RES | NIST-AI |
|---|---|---|---|---|---|---|---|---|---|
| Transparency Obligations | ● | ◐ | ◐ | ⚠ | ● | ● | ● | ◐ | ● |
| Foundation Models / GPAI | ● | ● | ◐ | ● | ● | ◐ | ◐ | · | ● |
| Deepfakes / Synthetic Content | ● | ● | · | ● | ● | · | · | ◐ | ◐ |
| Individual Redress | ● | · | ◐ | ● | · | ● | ● | · | ◐ |
| AI in Criminal Justice | ● | ● | ◐ | · | · | · | ● | · | · |
| Compute-Threshold Reporting | ● | ● | · | · | · | · | · | · | · |
| Sovereign AI Doctrine | · | ● | · | ● | · | · | · | · | · |
| Biometric Identification | ● | ◐ | ◐ | · | · | · | ◐ | · | · |
| AI in Employment | ● | ◐ | ◐ | · | · | · | ◐ | · | · |
| AI in Healthcare | ● | ◐ | ◐ | · | · | · | · | · | · |
| AI in Education | ● | ◐ | · | · | · | · | · | ◐ | · |
| Training-Data Rights | ◐ | · | · | ● | · | · | ◐ | · | ◐ |
Browse by topic
Topics ranked by editorial salience — governance density + conflict density. Each article shows the full 9-instrument coverage matrix and surfaces silent regimes as gap signals.
transparency
Transparency Obligations
Disclosure of training data, model cards, system-card requirements.
foundation_models
Foundation Models / GPAI
Obligations specific to general-purpose / foundation models above certain capability thresholds.
deepfakes
Deepfakes / Synthetic Content
AI-generated content disclosure, watermarking, election integrity protections.
redress
Individual Redress
Right to explanation, appeal mechanisms, complaint channels.
criminal_justice
AI in Criminal Justice
Predictive policing, risk assessment, sentencing assistance.
compute_reporting
Compute-Threshold Reporting
Mandatory reporting based on training-compute or capability thresholds.
sovereign_ai
Sovereign AI Doctrine
Domestic-compute, export controls, jurisdiction-bound model deployment.
biometric_id
Biometric Identification
Real-time and post-hoc biometric identification in public spaces.
employment
AI in Employment
Hiring, workplace monitoring, automated decisions in employment contexts.
healthcare
AI in Healthcare
Clinical decision support, medical devices, diagnostic AI.
education
AI in Education
Automated grading, proctoring, student-data analytics.
training_data
Training-Data Rights
Copyright, consent, text-and-data-mining exceptions.
Browse by instrument
9 major AI-governance instruments tracked across 8 jurisdictions / venues. Each article shows obligations, peer comparison, and citation-ready references.
United States
US-EO-14110
Executive Order 14110 on Safe, Secure, Trustworthy AI
Partially rescinded by EO 14179 (Jan 2025). Some §4 reporting persists via Defense Production Act + BIS interim rule.
NIST-AI-RMF
NIST AI Risk Management Framework
Voluntary. Four functions (Govern / Map / Measure / Manage). GenAI Profile (NIST AI 600-1) added 2024 for GPAI-specific guidance.
Browse capability benchmarks
10 public benchmarks tracked. Each article shows the methodology, per-model leaderboard, and contamination-risk guidance.
SWE-BENCH-VER · 2024
SWE-bench Verified
Solve real-world GitHub issues from 12 popular Python repos. The 'Verified' subset is human-validated to remove ambiguity and have working tests.
Contamination risk: medium
MMLU · 2020
MMLU
Massive Multitask Language Understanding — 57-subject multiple-choice covering humanities, STEM, social sciences, professional/legal.
Contamination risk: high
MMLU-PRO · 2024
MMLU-Pro
Successor to MMLU with 10-option multiple-choice (up from 4), more reasoning-focused tasks, and removed leaky / ambiguous items.
Contamination risk: medium
GPQA-DIAMOND · 2023
GPQA Diamond
Graduate-level Google-Proof Q&A in biology, chemistry, physics. 'Diamond' subset is the 198 hardest items.
Contamination risk: low
ARC-AGI-V2 · 2024
ARC-AGI v2
Abstract reasoning over visual grids. Each task requires inferring the transformation rule from 2-3 examples.
Contamination risk: low
HUMANEVAL · 2021
HumanEval
164 hand-written Python programming problems. Generate a function that passes provided unit tests.
Contamination risk: high
MATH · 2021
MATH (Hendrycks)
12,500 competition-math problems from AMC, AIME, etc. Evaluates step-by-step reasoning + final-answer accuracy.
Contamination risk: medium
AIME-2024 · 2024
AIME 2024
30 problems from the 2024 American Invitational Mathematics Examination — high-school competition math.
Contamination risk: low
HLE · 2025
Humanity's Last Exam
3,000+ frontier-difficulty expert-curated questions across all academic disciplines. Designed to remain unsaturated through 2026+.
Contamination risk: low
FRONTIER-MATH · 2024
FrontierMath
Hundreds of original research-mathematician-curated math problems requiring deep reasoning. Held-out evaluation only.
Contamination risk: low
Browse AI-governance concepts
Glossary articles for terms that recur across instruments — frontier-tier, systemic risk, ASL-3, compute thresholds. Each concept article links to the instruments that use the term.
frontier-tier · risk class
Frontier-Tier AI
A categorical classification of AI models above certain capability or compute thresholds, indicating heightened regulatory scrutiny.
asl-3 · safety
AI Safety Level 3 (ASL-3)
A capability-based risk tier in Anthropic's Responsible Scaling Policy denoting models with the potential to substantially uplift CBRN attack capabilities or autonomous AI replication.
systemic-risk · risk class
Systemic Risk (AI)
A regulatory designation indicating that a general-purpose AI model poses risks of significant scale or scope across the EU internal market, triggering Article 55 obligations under the EU AI Act.
designated-systemic · risk class
Designated Systemic-Risk Model
A general-purpose AI model that has been formally designated by the EU AI Office under Article 51(1)(b) as posing systemic risk, regardless of whether it meets the presumption thresholds.
compute-threshold · compute
Compute Threshold (AI Governance)
A regulatory trigger expressed as floating-point operations (FLOPs) consumed during model training, above which specific reporting, evaluation, or governance obligations attach.
Want more than the map?
Policy Window's analyst workbench layers paradigm runs (NPF · MSF · IAD · PE · Diffusion), substantive lenses (compute governance · international governance · alignment-layer · capability evals), and per-claim citation provenance on top of this wiki. Free for researchers, civil society, and verified policymakers.
Sign up freeCatalog refreshed at .