In September 2023, the G7 published draft Guiding Principles for Organisations Developing Advance AI Systems. Advanced AI Systems are stated to include foundation models and generative AI, and organisations include academia, civil society and the public and private sectors. At the end of October, the G7 leaders adopted the Principles.

The non-exhaustive list of eleven principles covers areas such as risk mitigation during development and deployment, identifying and mitigating post-deployment vulnerabilities and misuse, transparency and information sharing (including limitations and domains of appropriate or inappropriate use), AI governance and risk management, security controls, content authentication and provenance mechanisms (e.g. watermarking), prioritising research into AI risks and mitigations, development of AI to address the world’s greatest challenges (such as the climate crisis), development and adoption of international technical standards, and implementation of data input measures and protections for personal data and IP.

The Principles are followed by brief practical guidance such as listing the types of risks that should be considered during AI development (e.g. use of the AI for biological or chemical weapons development, use in cyber-attacks, capacity to control physical systems or to give rise to harmful bias, discrimination or disinformation).

In its announcement, the Commission states that the final version of the Principles and Code were informed by the results of the Commission’s stakeholder survey which it conducted earlier in October. It is expected that organisations developing AI systems will sign up to the Code in due course.

For more information, click here.