Those eleven points sit in the International Guiding Principles for Advanced AI Systems, one of two main outputs of the Hiroshima AI Process. The other is a voluntary International Code of Conduct that turns the principles into operational guidance for developers. Named for the Japanese city where G7 leaders met in May 2023, the process was an attempt by the world's leading industrial democracies (the United States, the United Kingdom, France, Germany, Italy, Canada, Japan, and the European Union) to set shared expectations while binding regulation was still years away.
What the Hiroshima Process produced
Unlike a bare political declaration, Hiroshima descended to practices. Companies and regulators received a common reference for what responsible frontier development should involve among the states that host most of the industry. In late 2023, it was the most detailed internationally agreed statement of that kind among leading democracies.
The Code of Conduct does not create legal duties. Organizations follow it voluntarily. Its force is reputational and political: a checklist others can cite when a lab's behavior falls short, and a floor that domestic rules can later raise.
The strength of the club model
The G7 format is fast and coherent. A small group of aligned democracies that already cooperate can reach agreement far more quickly than a universal negotiation, and can set standards that carry weight because members host much of the frontier industry. Many international regimes begin this way: a core group writes the standard, then others converge over time.
Hiroshima's outputs also fed the wider ecosystem. They informed OECD work and lined up with the direction later taken at Bletchley and Seoul. As a way to plant early norms among the states that matter most for AI development, the club model is efficient and still underrated.
The ceiling of the club model
- It excludes China. The most significant AI developer outside the G7 is not in the room. A code that binds only democracies cannot constrain the full field of frontier development, and cannot become universal governance on its own.
- It is voluntary. Like the Seoul company pledges, the code has no verification and no enforcement. It describes good practice; it does not compel it.
- It is a floor, not a ceiling. Guiding principles set minimum expectations. Binding thresholds, independent monitoring, and consequences for violation stay for harder processes.
The folk objection is that any club short of a UN treaty is empty symbolism. Concede the limit, then keep the use: clubs draft the rulebook quickly. A rulebook the other main player never signs, with no one checking and no penalty for ignoring it, remains a first draft. The draft still matters if someone later widens and hardens it.
Where it fits in the larger picture
Hiroshima is one of several parallel tracks (the summit series, the OECD, the UN) through which early AI norms are being assembled. Its contribution is a detailed, agreed statement of responsible practice among leading democracies, achieved by using existing trust. Its limit is structural: a like-minded club cannot deliver universal, binding, verified governance alone, because the actors it excludes are exactly the ones a comprehensive regime must include.
The realistic role of the G7 model is to set the standard and build momentum that a broader agreement can later formalize. That starting point only pays off if it is widened and given teeth. For catastrophic frontier risk, teeth means hard law: verified obligations, not a permanent code of voluntary best practice. See also how minilateral clubs fit a wider strategy.