G7 Hiroshima AI Process
In plain wordsAn effort begun by the G7 group of wealthy democracies in Japan in 2023 that sets voluntary rules for companies building the most advanced AI, which now report publicly on how they follow them.
An initiative of the Group of Seven (G7) on governing advanced AI, especially generative AI and foundation modelsfoundation model, n.: A very large AI model, built once at great cost from vast amounts of text, images or code gathered mostly from the internet, that many products then reuse as their starting point, from chatbots to coding tools.. Its core texts, released on 30 October 2023, are eleven International Guiding Principles for organizations developing advanced AI systems and a matching International Code of Conduct. Their commitments fall under four heads: managing risk before and after release, through testing such as red teamingred teaming, n.: Deliberately attacking your own AI system — trying to make it misbehave, leak secrets or help with harm — so that weaknesses are found before real adversaries or users find them. and watching for misuse once a system is deployed; openness, through public reports on capabilities and limits, incident sharing and published AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes. policies; protection, through security controls, content provenance tools such as watermarkingAI watermarking, n.: Hiding an invisible signal in AI-generated text, images, audio or video so that software can later tell the content was made by AI — like a banknote's watermark, but undetectable to the eye. and safeguards for personal data and intellectual property; and steering research and standards toward safety, societal benefit and international interoperability. With an OECD-led project track and guidance for all AI actors, they form the Comprehensive Policy Framework endorsed by G7 leaders in December 2023. Since 2025, organizations have reported on their adherence through a voluntary framework hosted by the OECD.
The lab's Hiroshima AI Process report describes its red-teaming and incident-sharing practices.
Origin
Launched at the G7 summit in Hiroshima, Japan, in May 2023 under Japan's presidency, months after ChatGPT's release. A Friends Group of non-G7 supporters formed in 2024, and the OECD reporting framework launched in February 2025, with the first organizational reports published that April.
Usage note
Voluntary and principle-based; often described as a bridge between the binding EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models. and the voluntary approaches of the United States, the United Kingdom and Japan. Several of its commitments reappear in the General-Purpose AI Code of PracticeGeneral-Purpose AI Code of Practice, prop. n.: A voluntary rulebook, published by the EU in 2025, that makers of large AI models can sign to show they follow the EU AI Act's rules on transparency, copyright and safety..
Related
- OECD AI PrinciplesOECD AI Principles, prop. n.: Guidelines for trustworthy AI that governments agreed through the OECD in 2019, the first such standard agreed between governments; many later laws and agreements borrowed from them.
- Frontier AI Safety CommitmentsFrontier AI Safety Commitments, prop. n.: Promises made in 2024 by major AI companies to publish safety plans for their most advanced models, including the danger points at which they would stop if the risks could not be controlled.
- AI governanceAI governance, n.: The rules, institutions and habits that shape how AI is built and used — from laws and treaties to technical standards and a company's own review processes.
- General-Purpose AI Code of PracticeGeneral-Purpose AI Code of Practice, prop. n.: A voluntary rulebook, published by the EU in 2025, that makers of large AI models can sign to show they follow the EU AI Act's rules on transparency, copyright and safety.
- red teamingred teaming, n.: Deliberately attacking your own AI system — trying to make it misbehave, leak secrets or help with harm — so that weaknesses are found before real adversaries or users find them.
- frontier AIfrontier AI, n.: The most capable general-purpose AI systems of the moment — those at or beyond the leading edge — which governments single out for extra safety attention.
Sources
- G7 (2023). Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems. 30 October 2023.
Mentioned in
OECD AI PrinciplesOECD AI Principles, prop. n.: Guidelines for trustworthy AI that governments agreed through the OECD in 2019, the first such standard agreed between governments; many later laws and agreements borrowed from them., AI Promotion ActAI Promotion Act, prop. n.: Japan's 2025 national AI law, which sets up a government AI headquarters and strategy to encourage AI development and use, relying on cooperation and guidance rather than fines., Global Dialogue on AI GovernanceGlobal Dialogue on AI Governance, prop. n.: A yearly United Nations meeting, launched in 2025, where all countries — joined by companies, researchers and civil society — discuss how to cooperate on governing AI., Global Partnership on Artificial IntelligenceGlobal Partnership on Artificial Intelligence, prop. n.: An international group of countries, launched in 2020 and now run together with the OECD, that brings governments and experts together to put trustworthy AI into practice., World AI Cooperation OrganizationWorld AI Cooperation Organization, prop. n.: An international organization for cooperation on AI, based in Shanghai, set up in 2026 by an agreement among 29 countries on China's initiative.