Guidance for the New Global Dialogue on AI Governance
This policy brief proposes an integrated, sequenced, United Nations (UN)-anchored roadmap for the new Annual Global Dialogue on AI Governance.
Its purpose is to help identify the key priorities the Global Dialogue needs to address, translate them into actionable design by outlining the contours of a Global AI Governance Roadmap (thereafter known as the “Roadmap”), and design the process required to deliver it.
In doing so, the brief aims for two goals:
- Structure a multistakeholder process that is practical, transparent, and predictable.
- Support implementation of the Pact for the Future and the Global Digital Compact.
Guidance for the New Global Dialogue on AI Governance Roadmap
Pillar I: Managing AI Risks
Outcome 1: Shared scientific and safety capacity for frontier models.
Annex I: Global Frontier AI Evaluation Framework
A common testing and evaluation framework co-developed with IISP-AI and regulators; shared risk categories; disclosure thresholds; baseline transparency requirements.
Outcome 2: Ethical and democratic safeguards for information ecosystems.
Annex II: Declaration on AI and Information Integrity
A negotiated declaration establishing expectations for content authenticity, electoral safeguards, platform obligations, and human oversight in communicative environments.
Pillar II: Distributing AI Rewards
Outcome 3: Equitable access to compute, data, skills, and innovation capacity.
Annex III: AI Infrastructure Access Framework
A global framework linking regional capacity hubs, shared compute access, multilingual datasets, and a voluntary trust fund or credit mechanism.
Outcome 4: Responsible and accountable AI in public administration.
Annex IV: Responsible Public-Sector AI Guidelines
A set of minimum operational standards for procurement, transparency, auditability, and human oversight in government AI systems, including for private contractors.
Pillar III: Aligning AI Rules
Outcome 5: Reduced regulatory fragmentation and improved baseline coherence.
Annex V: Regulatory Interoperability Mechanism
A mechanism defining minimum global documentation and disclosure baselines; templates for mutual recognition; and a UN facilitation platform.
Outcome 6: A coherent global institutional architecture for AI governance.
Annex VI: Institutional Coherence Options Paper
A mapping of existing bodies and three coordination models (council, inter-agency mechanism, observatory) with functions, reporting lines, and options for long-term architecture.
