Most AI governance today targets narrow AI — systems optimized for specific tasks. AGI is different: by definition, it could match or exceed human performance across most cognitive domains.

That creates three hard realities:

• Uncertain timelines, real momentum — credible experts disagree on timing, but capabilities are moving toward more general, autonomous systems
• Global externalities — an AGI deployed by one actor could affect everyone, from financial stability to information integrity and strategic deterrence.
• No dedicated global regime —unlike nuclear energy or civil aviation, there is no agreed international framework specifically for AGI.

Based on the Millennium Project’s study, Jerome Glenn was asked by the Council of Presidents of the United Nations General Assembly to form and chair the High-Level Expert Panel on AGI.  The participants were Yoshua Bengio (Canada), Stuart Russell (UK), Lan Xue (China), Mariana Todorova (Bulgaria), José Villalobos (Costa Rica), Jaan Tallinn (Estonia), Joon Ho Kwak (Republic of Korea), and Renan Araujo (Brazil), in their personal capacities. The panel’s report is freely available in several languages.

AGI governance will fail if it is designed only by a small circle of experts and powerful states. To help prevent the greatest concentration of power, The Millennium Project combines high-level expertise with global participatory methods.

• Real-Time Delphi (RTD) — structured expert surveys that remain open for weeks, allowing participants to revise their judgements as they see others’ arguments and new evidence.
• Polis Conversations — large-scale online conversations where participants agree or disagree with short statements, revealing areas of consensus and division in real time.
• Node-driven dialogues — facilitated discussions run by our 70+ Nodes around the world, translating AGI governance debates into local contexts.

We document these approaches with futures research methodologies and share them with partners who want to run their own AGI-related foresight processes.

We work with institutions that need to understand what AGI could mean for their mandates and how to prepare.

We can support you with:

• Confidential briefings on the UN AGI report, Delphi results, and governance options.
• Scenario workshops on how AGI could affect your sector or region.
• Support for national or regional AGI strategies, including stakeholder consultations using RTD and Polis.
• Input into international norm-building and multi-stakeholder processes.

Our online Commons hosts ongoing discussions on AGI governance, including commentary on the UN report and case studies from different regions.

We use Polis to map where people agree and disagree on key AGI governance questions — from licensing to global institutions.

Levels of AGI for Operationalizing Progress on the Path to AGI

OECD AI Capability Indicators