Hiroshima AI Process

Hiroshima AI Process is an international initiative launched by G7 leaders in May 2023 to promote responsible AI development globally, with a focus on generative AI and foundation models.

Key Outcomes:

• International Guiding Principles: 11 non-binding principles for organizations developing advanced AI systems
• International Code of Conduct: Voluntary code for organizations developing advanced AI systems
• Project-Based Cooperation: Practical collaboration among G7 members on AI governance

The 11 Guiding Principles:

1. Take appropriate measures to identify, evaluate, and mitigate risks
2. Identify and mitigate vulnerabilities, incidents, and misuse after deployment
3. Publicly report capabilities, limitations, and use cases
4. Work toward responsible information sharing among organizations
5. Develop AI governance and risk management policies
6. Invest in robust security controls
7. Develop content authentication and provenance mechanisms
8. Prioritize research to mitigate societal risks
9. Prioritize development to address global challenges
10. Advance international technical standards
11. Implement data input measures and data protection

Relationship to EU AI Act:

• AI Act Article 53 allows compliance with international codes of conduct (like Hiroshima) to demonstrate compliance with certain GPAI obligations
• Particularly relevant for GPAI providers outside the EU
• Supports international regulatory convergence

Implementation: The G7 encourages all AI actors to voluntarily commit to these principles and the code of conduct, while individual countries develop binding regulations.