Generative AI: Global Governance and the risk-based approach

The emergence of generative AI has sparked both enthusiasm and concern, especially because there is a currently a lack of know how of the technology itself. Given the significant private sector involvement in AI and the global nature of AI regulation, there is a growing need for comprehensive, technology-neutral, multi-stakeholder-driven regulatory frameworks.

In addition to this need, there is a growing consensus on exploring potential risks from generative AI, while sensationalist media are driving the divide between existential and immediate concerns. This policy report aims to reframe the political discourse around regulating foundational technologies like generative AI, offering practical policy approaches and recommendations for global convergence. The “risk-based approach” from the EU AI Act is an illustrative example that policymakers, including those in the G7 Hiroshima AI Process, can adopt to assess any potential foundation model risks. This perspective is also relevant to initiatives like the Global Partnership on Artificial Intelligence (GPAI) and the G20, led by countries such as India and Brazil.

This policy report comprises three main sections: one addressing risks arising from the use of generative AI, another discussing risk mitigation measures for the risks that were surfaced, and a final section charting a path for global governance of generative AI. It concludes with concrete policy recommendations for regulatory convergence through evidence-based, technology-neutral, multi-stakeholder, resilient policymaking. This aligns with the goal of CERRE’s Global Governance for the Digital Ecosystems project (GGDE) to  promote regulatory convergence globally and ensuring co-existence when convergence is not feasible.

This report is the first in a series of upcoming GGDE policy notes to explore values, frameworks, and global convergence on AI regulation. This series will explore principles, human oversight, comprehensibility, accessibility, competition law, copyright, AI risk assessment, and international AI system transfer/use. The project will also examine the roles of EU institutions and other bodies in organising convergence.

The overarching goal is to provide a concrete path forward for G7+ policymakers, fostering a common understanding of “AI adequacy” to guide industry and citizens worldwide and ensuring progress on the Hiroshima AI Process. This report also provides evidence-based policy recommendations for stakeholders in these processes. CERRE has already submitted four specific policy recommendations within the framework of the European Commission “stakeholder survey” of October 2023, guiding the G7 Hiroshima AI Process:

• Adapt the “risk-based approach” from the EU AI Act for Generative AI in G7+ countries, focusing on “high-risk” general purpose AI model applications and developing industry risk assessment tools.
• Focus on developing tiered, yet limited regulatory obligations for high-risk use cases.
• Embrace a collaborative, multi-stakeholder approach in AI governance efforts.
• Promote “AI adequacy” to enhance legal certainty and regulatory convergence.