Summary
- Came out of the Hiroshima AI Process, started by the G7 under Japan's presidency in May 2023. The Code was released with the G7 Leaders' Statement of 30 Oct 2023 and folded into the Comprehensive Policy Framework endorsed by G7 Leaders in Dec 2023.
- Aimed at organisations developing advanced AI systems (the most advanced foundation models and generative AI). The preamble calls it voluntary guidance, to be applied with a risk-based approach, and a living document built on the OECD AI Principles.
- The 11 actions, summarised: (1) assess and reduce risks before and during deployment; (2) find and fix vulnerabilities, incidents and misuse after release; (3) publicly report capabilities, limits and suitable uses; (4) share information and report incidents responsibly; (5) set and disclose AI governance and risk policies; (6) invest in security, including insider threats; (7) build content authentication and provenance such as watermarking where feasible; (8) prioritise research on societal, safety and security risks; (9) direct AI at global challenges like climate, health and education; (10) help develop and adopt international technical standards; (11) protect data inputs, personal data and intellectual property.
- Reporting Framework: the OECD launched it on 7 Feb 2025 so companies can report comparably on how they apply the Code. 19 companies' first responses were published in April 2025 (per the Hiroshima AI Process site). Version 2.0, launched 28 May 2026, is streamlined for SMEs and deployers; responses from 31 companies were published on 2 Oct 2026.
Who it applies to
- Organisations that choose to sign up, mainly developers of frontier models. Version 2.0 widens the reporting form to deployers and smaller firms.
- India is a member of the HAIP Friends Group (supporting countries list, as of Aug 2026). Infosys is listed in the Partners Community (as of Oct 2026). Membership is a statement of support, not a legal commitment.
- Not extraterritorial.
Key dates
- Hiroshima AI Process launched after the G7 Hiroshima Summit
- G7 Leaders' Statement on the Hiroshima AI Process; Guiding Principles and Code of Conduct released
- G7 Digital and Tech Ministers agree the Comprehensive Policy Framework
- G7 Leaders endorse the Comprehensive Policy Framework
- HAIP Friends Group of supporting countries launched
- OECD launches the HAIP Reporting Framework
- Deadline for inaugural reports; rolling submissions after that
- OECD launches Reporting Framework 2.0, streamlined for SMEs and deployers
- Reporting Framework 2.0 responses from 31 companies published
Full date record
| Date | Event |
|---|---|
| 2023-05 | Hiroshima AI Process launched |
| 2023-10-30 | G7 Leaders' Statement; Code of Conduct released |
| 2023-12-06 | G7 Leaders endorse the Comprehensive Policy Framework |
| 2024-05 | Friends Group launched |
| 2025-02-07 | OECD Reporting Framework launched |
| 2025-04-15 | Inaugural report deadline |
| 2026-05-28 | Reporting Framework 2.0 launched |
| 2026-10-02 | 2.0 responses from 31 companies published |
Penalties and enforcement
- None. Reports are voluntary and published on the OECD transparency portal; the only pressure is reputational.
What it means for an Indian company
- Frontier model builders (for example Indian foundation model projects) can use the 11 actions as a ready checklist and consider filing an OECD report to show maturity to global buyers.
- SaaS firms that deploy AI can now use version 2.0 as a structured self-disclosure.
- The actions map well to controls: SC-AI-018 Robustness and adversarial testing (red teaming) (1), SC-AI-031 AI incident management and regulator reporting (2, 4), SC-AI-014 Model documentation (model cards, system cards) and SC-AI-025 Public AI transparency statement and notices (3, 5), SC-AI-029 AI-specific security controls (prompt injection, model abuse) (6), SC-AI-023 Label AI-generated and synthetic content (7), SC-AI-010 Training and input data governance (provenance, quality) and SC-AI-011 Lawful basis and consent for personal data used in AI (11).
- Action 7 on provenance aligns with India's own push on synthetic content labelling (see IT Rules Amendment 2026 (Synthetic Content)).
Obligations
Full obligation breakdown coming
We break the most requested policies down into numbered obligations mapped to controls. Leave your details and we’ll tell you when G7 Hiroshima AI Code of Conduct is done.
Glossary
- Hiroshima AI Process (HAIP): the G7 initiative on generative and advanced AI started in 2023.
- Advanced AI system: in the Code, the most capable foundation models and generative AI.
- Reporting Framework: the OECD questionnaire where organisations describe how they apply the Code.
- Friends Group: countries that support the spirit of the Hiroshima AI Process.
- Content provenance: technical signals showing where a piece of content came from and whether AI made it.
Key takeaways
- Voluntary 11-action code for frontier AI developers, released 30 Oct 2023.
- OECD reporting launched 7 Feb 2025; version 2.0 on 28 May 2026 opens it to SMEs and deployers.
- India supports it through the Friends Group but has no legal duty under it.
Official sources
- Hiroshima AI Process: Documents of Achievement (includes the Code text), accessed 2026-10-19
- Hiroshima AI Process: Reporting Framework page, accessed 2026-10-19
- Hiroshima AI Process: Friends Group membership, accessed 2026-10-19
- OECD press release, 7 Feb 2025: framework to monitor the Code, accessed 2026-10-19
- OECD press release, 28 May 2026: Reporting Framework 2.0, accessed 2026-10-19
This is general information, not legal advice. Last verified 19 Oct 2026 against the official source. Laws change; confirm the current position with qualified counsel before acting.