When the EU, the US, and Japan can't agree on a single AI law, the G7 found the alternative: a code of conduct nobody is legally required to sign, that nearly every relevant frontier lab signed anyway.
The Hiroshima AI Process (HAIP) was born at the G7's Hiroshima Summit in May 2023, as a response to the accelerating proliferation of large-scale generative models and frontier capabilities. The International Code of Conduct for Organizations Developing Advanced AI Systems was formally adopted on October 30, 2023.
The code sets out eleven guiding principles with corresponding actions, focused specifically on developers of the most advanced AI systems — frontier models and general-purpose foundation models. They cover areas like identifying and mitigating risks across the lifecycle, publishing reports on capabilities and limitations, investing in cybersecurity, and prioritizing the development of systems that address humanity's most pressing challenges.
The code is explicitly voluntary and has no legal force in any jurisdiction. Even so, it functions as a soft-law commitment mechanism designed to shape responsible AI development globally — its value doesn't come from forced enforcement, but from the fact that most relevant frontier labs (OpenAI, Anthropic, Google DeepMind, among others) signed it publicly, creating a de facto norm that later influences hard regulation in other jurisdictions.
Hiroshima Process documents are reviewed and updated as needed, drawing on multi-stakeholder consultation opportunities in response to accelerating technological development — it's one of the few forums where representatives of the major AI powers (the US, EU, Japan, UK, and the rest of the G7) coordinate directly on shared principles, even when their national laws diverge enormously (compare Japan's no-fines approach with Europe's sanctions regime).
If your company evaluates frontier model providers as part of a due-diligence process, checking whether a provider signed and complies with the Hiroshima Code is a reasonable signal of commitment to responsible practices — it doesn't replace compliance with specific hard laws like the EU AI Act, but it's a useful reference when evaluating providers in jurisdictions where applicable hard regulation doesn't exist yet.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel