Three instruments, one job
Lesson 1 of 5 in Global Soft Law: OECD, UNESCO, and the G7 Hiroshima Process.
No fine has ever been issued under the OECD AI Principles. No court has ever applied the UNESCO Recommendation. No regulator enforces the G7’s code of conduct. And yet these three documents shaped more binding AI law than any statute you will study — because they built the shared vocabulary, evidence base, and political consensus that legislatures later wrote into force.
You already know the theory from soft law 101: non-binding instruments are where consensus forms before parliaments act. This module is about the machinery — the observatories, adherence lists, assessment methodologies, and reporting frameworks that make three pieces of paper behave like institutions. The principles themselves — fairness, transparency, accountability and the rest — are taught properly in the ethics module; here you learn who runs the apparatus and why it matters.
| Instrument | Form | Adopted | Who signed up | Structure | The machinery behind it |
|---|---|---|---|---|---|
OECD AI Principles | Recommendation of the OECD Council (OECD/LEGAL/0449) | May 2019; revised May 2024 | 47 adherents — all OECD members plus non-members including Brazil, Singapore, and Ukraine, and the EU; echoed by the G20 | 5 values-based principles + 5 recommendations to policymakers | OECD.AI Policy Observatory, AI Incidents Monitor, expert networks, the integrated GPAI partnership |
UNESCO Recommendation on the Ethics of AI | Recommendation of the General Conference | November 2021 | All UNESCO member states — including China and Russia; the broadest AI instrument on Earth | 4 core values + 10 principles + 11 policy action areas | Readiness Assessment Methodology (RAM), Ethical Impact Assessment (EIA), Global AI Ethics and Governance Observatory, Global Forums |
G7 Hiroshima AI Process | G7 leaders’ statement + code of conduct | October 2023; reporting framework February 2025 | G7 members, then a Friends Group of roughly 50 countries; companies join by filing reports | 11 guiding principles + an International Code of Conduct for advanced AI developers | OECD-hosted voluntary transparency reporting framework — the first common disclosure channel for frontier labs |
Notice the pattern in the last column. Each body compensated for the absence of enforcement with a different kind of infrastructure: the OECD built measurement, UNESCO built assessment, and the G7 built disclosure. Keep that trio in mind — every lesson in this module unpacks one of them.
One more orientation point: these three do not compete so much as they layer. The OECD text speaks to governments of industrialized democracies; UNESCO extends a broader ethics baseline to every UN member state, including states the OECD will never include; the Hiroshima Process speaks directly to the dozen-odd companies building frontier models. Same principles vocabulary, three different audiences.
The soft-law arc — from privacy guidelines to AI principles
- 1980-09-23 — OECD Privacy Guidelines:
The first international data-protection framework — the template for cross-border governance of information technology, four decades before the same body wrote AI principles.
- 2019-05-22 — OECD AI Principles adopted:
The first intergovernmental AI standard — 40+ adherents including the US and (via G20) China. Its AI-system definition becomes the shared vocabulary of the EU AI Act and US law.
- 2021-11-23 — UNESCO Recommendation on the Ethics of AI:
Adopted by 193 countries — the broadest AI ethics instrument on Earth, with readiness-assessment machinery for implementation.
- 2024-05-03 — OECD AI Principles updated:
The 2019 principles get a generative-AI refresh — including the revised AI-system definition that laws worldwide now cite.
Key terms: soft law, hard law, OECD AI Principles, UNESCO Recommendation on the Ethics of AI, Hiroshima AI Process
Interactive checkpoint quiz (2 questions) — open this page in a browser to take it.