Every AI Policy Ever · open in the interactive register

Published (non-binding, current)Non-binding · Voluntary code / commitments

Hiroshima AI Process (HAIP) Reporting Framework for the International Code of Conduct for Organizations Developing Advanced AI Systems

Record OECD-02

Issued by
OECD, for the G7 Hiroshima AI Process
Jurisdiction
OECD / G7
Level · region
International · Global
Published / adopted
2025-02-07
Last updated
2026-05-28
Language
English

Summary

The OECD's framework for organisations developing, deploying or providing advanced AI systems to report publicly on their alignment with the G7 Hiroshima AI Process International Code of Conduct. Built around the Code's 11 actions, its questionnaire covers risk identification and evaluation, risk management and information security, transparency on advanced AI systems, governance and incident management, content authentication and provenance, and research on AI safety. Participation is voluntary and reports are published in full on OECD.AI.

Safety topics

transparency & labellinggeneral-purpose / frontier AIaccountability & liabilityincident reportinginternational cooperation

Status notes

Launched on 7 February 2025 as an outcome of the G7 Hiroshima AI Process, after the G7 Digital and Tech Working Group under the Italian Presidency, following the Trento Declaration, asked the OECD to develop mechanisms for monitoring voluntary adoption of the Code. The OECD counts 25 reports in the first round, analysed in its 2025 report 'How are AI developers managing risks?'; MIC Japan's HAIP site says the first responses, from 19 companies, were published in April 2025. Version 2.0, launched on 28 May 2026 at a Tech7 event in Paris on the margins of the G7 Digital and Tech Ministerial under the French Presidency, distinguishes model developers, application developers and deployers and is simplified for small and medium enterprises; 31 reports under it were published on 2 October 2026. Organisations must be based in an OECD member, a GPAI member or an Adherent to the OECD Recommendation on AI.

Sources

Open the official source https://oecd.ai/en/transparency/overview

Secondary source: https://www.soumu.go.jp/hiroshimaaiprocess/en/report.html

Local copy

Archived. Last attempt 2026-10-18 06:47:42 UTC

Official source (third-party copy, not published) archive/OECD-02/2026-10-17/official.html
Retrieved 2026-10-17 14:28:21 UTC
SHA-256 46d38f702a5553d84e2db9b78460ede78b258752faaa28ac3c1f10ddd9fe7197
Secondary source (third-party copy, not published) archive/OECD-02/2026-10-17/secondary.html
Retrieved 2026-10-17 14:28:28 UTC
SHA-256 348bf2a18e4c8f013bb6cb7d44954aa55e2b1bce00968e7aac6efabeedb56bc4

Verification

A
The instrument's own official text, or an official register, was stored and confirms the title, date and status. checked 2026-10-18 · official-url-fetched
Refresh of 18 Oct 2026. Stored for this record: the OECD's overview page of the Reporting Framework on OECD.AI (the official URL, 17 Oct 2026), the framework's own page, which sets who may take part and how reports are submitted and published and gives the launch on 7 February 2025, the G7 mandate under the Trento Declaration, the 25 reports of the first round and version 2.0; and MIC Japan's Hiroshima AI Process page on the framework (secondary source), which gives the same launch date, the first responses from 19 companies in April 2025 and the 31 reports under version 2.0 published on 2 October 2026. Stored in this run: the blank questionnaire of version 2.0 (Word file, oecd.ai), headed 'Hiroshima AI Process Voluntary Reporting Framework V2.0', whose seven sections, each tied to actions of the Code of Conduct, confirm the summary's list of what the questionnaire covers. The OECD.AI 'about' and 'reports' pages and the OECD.AI announcement of version 2.0 (launched 28 May 2026; the announcement is dated 29 May 2026) were stored by the recheck of 17 Oct 2026. The February 2025 OECD press release on oecd.org refuses the archiver (HTTP 403).
Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 5 field(s) supported.

History

  1. 2026-10-18related: [{"slug": "hiroshima-process-code-of-conduct", "id": "G7-02"}, {"slug": "g7-mini → [{"slug": "hiroshima-process-code-of-conduct", "id": "G7-02"}, {"slug": "g7-mini — two-way link added: G7-03, G7-05, INT-20 link(s) to OECD-022026-10-18-fix-related-links · Claude Code (claude-opus-5-5)
  2. 2026-10-18Re-verified against its sources, no change — pipeline run 2026-10-18-refresh-level-rule-1 (refresh), candidate OECD-022026-10-18-refresh-level-rule-1 · Claude Code (claude-opus-5-5)
  3. 2026-10-17notes: Launched 7 Feb 2025. Reports from 19 companies in the first round were published → Launched on 7 February 2025 as an outcome of the G7 Hiroshima AI Process, after — the G7 mandate, both first-round counts (OECD 25 reports, MIC 19 companies), the version 2.0 launch date and content and the eligibility rule now rest on stored OECD.AI pages2026-10-17-recheck-international-1 · Claude Code (claude-opus-5-5)
  4. 2026-10-17safety_topics: ["transparency & labelling", "general-purpose / frontier AI", "accountability & → ["transparency & labelling", "general-purpose / frontier AI", "accountability & — 'incident reporting' added: the framework asks about incident reporting and incident management (about page; Section 4 of the questionnaire)2026-10-17-recheck-international-1 · Claude Code (claude-opus-5-5)
  5. 2026-10-17summary: The first global framework for companies to report publicly on what they do to p → The OECD's framework for organisations developing, deploying or providing advanc — rewritten from the OECD's own description: organisations rather than companies, the Code's 11 actions and the questionnaire's sections; 'first global framework' dropped (the OECD says 'the only international framework', a claim about itself)2026-10-17-recheck-international-1 · Claude Code (claude-opus-5-5)
  6. 2026-10-17official_url: https://www.oecd.org/en/about/news/press-releases/2025/02/oecd-launches-global-f → https://oecd.ai/en/transparency/overview — the OECD's own page of the framework on OECD.AI, stored in this run; the February 2025 press release on oecd.org has refused the archiver in every attempt since 11 Oct 20262026-10-17-recheck-international-1 · Claude Code (claude-opus-5-5)
  7. 2026-10-17date_last_updated: 2026-10-02 → 2026-05-28 — 2026-10-02 was the publication of version 2.0 reports, not a new version; version 2.0 itself was launched on 28 May 2026 (OECD.AI announcement published 29 May 2026)2026-10-17-recheck-international-1 · Claude Code (claude-opus-5-5)

Related records

G7 Hiroshima Code of Conduct G7 · Published (non-binding, current)

G7 Digital & Tech Ministerial Declaration 2026 (Paris) G7 · Published (non-binding, current)

OECD AI Principles OECD · Published (non-binding, current)

HAIP Comprehensive Policy Framework G7 · Published (non-binding, current)

G7 Leaders' Statement on AI for Prosperity (Kananaskis, 2025) G7 · Published (non-binding, current)

HAIP Friends Group Action Plan 2026 Hiroshima AI Process Friends Group · Published (non-binding, current)

Citation

OECD, for the G7 Hiroshima AI Process. Hiroshima AI Process (HAIP) Reporting Framework for the International Code of Conduct for Organizations Developing Advanced AI Systems. 2025-02-07. https://oecd.ai/en/transparency/overview. Status: Published (non-binding, current); verified 2026-10-18 (level A). Local copy archive/OECD-02/2026-10-17/official.html retrieved 2026-10-17 14:28:21 UTC, SHA-256 46d38f702a5553d84e2db9b78460ede78b258752faaa28ac3c1f10ddd9fe7197. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record OECD-02.

Improve this record

Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.