The CERN model for AGI research
"Image synthesis assisted by Grok Imagine Image, an AI partner within the Global Future Nexus ecosystem."
From particle physics to artificial general intelligence, the CERN model is emerging as the most compelling blueprint for collaborative, public-interest-driven AGI research—but translating the analogy into reality requires institutional ambition that current proposals have yet to match.
An Analogy with Power
When cognitive scientist Gary Marcus first coined the phrase “CERN for AI” at the 2017 AI for Good Summit, he invoked a powerful vision: cooperative, public-interest-driven development of transformative technology, modelled on the European Organisation for Nuclear Research. Since then, the analogy has gained remarkable traction. CERN, founded in Geneva in 1954 as a joint European project for fundamental, civilian, and open research in particle physics, has become the dominant metaphor for how the world might organise AGI research. The underlying logic is compelling: if collaborative multinational research could unlock the secrets of the subatomic world, why not the secrets of general intelligence?
Yet the gap between analogy and reality remains vast. As a RAND Corporation analysis observed, there are “numerous overlapping and competing descriptions of what functions a CERN for AI might serve”. The term risks becoming a convenient shorthand for ambition without substance.
Five Concepts, One Challenge
RAND researchers have identified five distinct concepts for a CERN for AI that capture different dimensions of the visions scholars have articulated:
A collaborative research institute focused on foundational AGI research, bringing together the world's brightest minds under one institutional roof.
A public option for foundation models —a trusted, transparent, and accountable alternative to private AI infrastructure, ensuring that advanced AI capabilities remain accessible to the public good rather than concentrated in a handful of corporations.
A testing and alignment facility where AGI systems could be rigorously evaluated in controlled environments before deployment—a “CERN for AGI” focused on autonomous simulation-based testing and alignment.
A governance mechanism providing international oversight of AGI development, fostering transparency and ensuring that AGI is developed safely and equitably.
A talent magnet that consolidates European AI research across fragmented hubs, creating the critical mass necessary to compete with Silicon Valley.
These concepts are not mutually exclusive, but they impose different institutional requirements. A testing facility demands different capabilities than a foundation model builder. A governance mechanism requires different legitimacy than a research institute. The ambiguity is not merely academic—it shapes what a CERN for AI could actually achieve.
The European Proposals: EDIRAS and Beyond
The most concrete proposal has emerged from Europe. In April 2024, the Group of Chief Scientific Advisors to the European Commission recommended establishing a European institute for AI in science—not as a single centralised organisation, but as a distributed institution they dubbed EDIRAS (European Distributed Institute for AI in Science).
The distributed model, the Advisors argued, “would bring many advantages and avoid some of the pitfalls of a single monolithic entity”. EDIRAS would provide fair access to massive high-performing computational power, sustainable cloud infrastructure, funding for AI research, and AI training programmes for researchers across all disciplines. It would ensure robust and ethical AI development aligned with European values including fairness, transparency, and inclusivity.
Building on this, the Centre for Future Generations has proposed a more ambitious blueprint: a treaty-based, independent, pan-European research institution with long-term political commitment, shared ownership by states, and deep trust in the scientific community. The proposed institution would bridge foundational research with real-world impact across four areas: fundamental research into AI interpretability and reliability; scaling promising approaches into cutting-edge systems; developing applications for pressing societal challenges; and advancing hardware foundations for responsible AI development.
The governance model relies on twin principles of transparency and accountability, implemented through a Member Representative Board that serves as guardian of the institution's mission, alongside independent expert boards ensuring work stays aligned with institutional values through public feedback loops. The estimated cost: €30-35 billion over the first three years.
The Criticism Gap
Yet the gap between proposal and reality has drawn sharp criticism. In January 2026, Morten Irgens and Holger Hoos of CAIRNE argued that current EU initiatives branded as a “CERN for AI” fall short of what the model truly entails. CERN succeeded, they noted, not because it funded individual projects, but because it was established as a treaty-based, independent institution with long-term political commitment and shared ownership by states.
By contrast, today's AI initiatives remain fragmented, time-limited, and constrained by short-term funding and governance structures. The authors cautioned that “reusing the CERN label without matching its defining characteristics risks weakening both the concept and Europe's credibility”.
Other critics have raised deeper concerns. Some argue that a centralised international group cannot align superintelligence any more than decentralised organisations can. Others contend that using the CERN label for AI research infrastructure is “misleading” because the current political circumstances are not comparable to those that facilitated CERN's establishment. As one observer noted, building a CERN for AI requires bringing together highly specialised and diverse expertise that currently does not exist within EuroHPC centres.
GFN's Role: A Distributed Alternative
Global Future Nexus is uniquely positioned to complement—and in some ways transcend—the CERN for AI model. GFN's distributed structure, spanning over 200 countries, embodies the collaborative, multinational ethos that CERN represents without requiring a single centralised institution. GFN's work on AGI identity, ethics frameworks, and sustainable integration offers practical infrastructure for the kind of international cooperation that CERN for AI proponents envision.
If the CERN model represents centralised excellence, GFN represents distributed stewardship—a recognition that AGI governance cannot be confined to a single laboratory, however ambitious. The challenge is not to choose between these models but to ensure they reinforce each other. As the Simon Institute for Longterm Governance concluded, the time to establish a CERN for AI-style institution is now—technology is moving fast and the geopolitical environment is challenging. But the institution must be worthy of the name.
The CERN model for AGI research remains a compelling vision. Whether it becomes reality depends on whether the world can translate ambition into institutional architecture—and whether that architecture can keep pace with the technology it seeks to govern.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
