G7 Leaders Launch Hiroshima Process for Generative AI — October 30, 2023
G7 leaders released guiding principles and a code of conduct for advanced AI developers under the Hiroshima AI Process.
The principles emphasise risk identification, security controls, transparency, incident reporting, and responsible governance for advanced foundation models. The voluntary code of conduct calls for risk assessments across the lifecycle, cybersecurity safeguards, red-teaming, and disclosure of capabilities and limitations.
- Lifecycle governance. Developers should implement risk management from design through deployment and ongoing monitoring.
- Transparency. The guidance encourages disclosure of AI-generated content, training data policies, and user education.
- Incident response. Providers are urged to share significant incidents with stakeholders and governments.
Vendors can align Hiroshima Process expectations with U.S. voluntary commitments and EU AI Act duties to harmonise compliance.