The Global Governance Void: Why the World Needs an International Body for Artificial Intelligence

PARIS — Several years ago, economist Mariana Mazzucato and I posed a deceptively simple question: Who is truly in charge if something goes violently, systemically wrong in the artificial intelligence ecosystem?

Today, that question remains painfully unanswered. Even after high-level deliberations, including a dedicated meeting by the United Nations Security Council to address the existential risks posed by rapidly advancing AI systems, the global community finds itself adrift. Artificial intelligence is already concentrating enormous economic power into the hands of a microscopic oligopoly of technology conglomerates while deepening existing socioeconomic inequalities. Yet, national governments largely lack the institutional frameworks, regulatory teeth, and technical expertise needed to govern it effectively.

To bridge this widening chasm, the international community must urgently establish a dedicated international AI body. Such an institution would be tasked with setting rigorous, binding standards and coordinating a unified, collective response to cross-border threats that no single nation can handle in isolation.

Main Facts

The rapid proliferation of generative artificial intelligence and frontier models has outpaced traditional legislative cycles, creating a profound governance deficit.

• Concentration of Economic Power: A handful of multinational technology firms control the computational infrastructure, proprietary datasets, and specialized talent required to train frontier AI models. This creates a winner-take-all dynamic that concentrates wealth and market dominance on an unprecedented scale.
• Deepening Inequalities: Within nations, displacement driven by algorithmic automation threatens labor markets, disproportionately impacting working-class and marginalized communities. Globally, the "AI divide" threatens to relegate developing economies to consumers of foreign technology rather than sovereign producers, exacerbating historical global disparities.
• The Governance Vacuum: While regional frameworks like the European Union’s Artificial Intelligence Act have made strides, regulatory approaches remain fragmented. There is no overarching global arbiter capable of auditing frontier algorithms, enforcing safety baselines, or preventing a regulatory "race to the bottom."
• Systemic and Existential Risks: From autonomous cyberattacks and the proliferation of weapons of mass destruction to algorithmic bias in critical infrastructure (such as healthcare and criminal justice), the hazards posed by advanced AI do not respect national borders.

Chronology of the Crisis: How We Reached the Brink

The trajectory of artificial intelligence governance over the past decade has been characterized by reactive policy-making chasing exponential technological breakthroughs.

• November 2022: The public release of OpenAI’s ChatGPT ignites a global generative AI boom, capturing public imagination and venture capital while catching regulators flat-footed.
• December 2022: Gabriela Ramos and Mariana Mazzucato publish a landmark critique warning that ethical AI requires robust state regulatory frameworks and deliberate capacity building, explicitly asking who would be held accountable when systems fail.
• Throughout 2023: Governments race to respond. The UK hosts the inaugural AI Safety Summit at Bletchley Park, resulting in the Bletchley Declaration, which emphasizes international cooperation on frontier AI risks. Simultaneously, the White House issues a sweeping Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence.
• Early 2024: The European Union formally adopts the AI Act, becoming the first major Western jurisdiction to establish a comprehensive, risk-based legal framework for artificial intelligence. However, critics note its enforcement relies heavily on national authorities with varying levels of technical literacy.
• Late 2024 – 2025: Major tech firms deploy increasingly autonomous agentic systems. Reports of algorithmic hallucinations in critical financial and healthcare systems multiply, while deepfakes disrupt democratic elections across dozens of countries.
• Late 2026: The UN Security Council convenes to discuss the existential security risks of AI. Despite high-level rhetoric, actionable global enforcement mechanisms remain entirely absent, underscoring the paralysis of traditional multilateral institutions in the face of private-sector technological dominance.

Supporting Data and Economic Realities

The argument for a centralized international AI body is underpinned by stark economic and technical data illustrating the scale of the imbalance between state capacity and corporate power.

• Compute Concentration: Training state-of-the-art frontier models requires clusters costing hundreds of millions, or billions, of dollars in specialized hardware (such as advanced GPUs and TPUs). According to industry analyses, over 80% of global frontier compute capacity is controlled by fewer than five private corporations.
• The Talent Drain: Public sector agencies routinely struggle to recruit and retain top-tier AI researchers. Private tech giants routinely offer compensation packages that dwarf government salaries, leaving regulatory bodies heavily reliant on industry self-reporting—effectively allowing corporations to mark their own homework.
• Global Disparities in AI Readiness: According to Oxford Insights’ Government AI Readiness Index, high-income nations dominate the upper tiers of infrastructure and governance preparedness, while developing nations in Sub-Saharan Africa and parts of Latin America score critically low, threatening to lock them out of the productivity gains of the Fourth Industrial Revolution.
• Environmental Toll: The training and inference phases of large language models consume vast quantities of energy and water. Data center expansions threaten national climate targets, an externality that current market dynamics fail to price or regulate adequately.

Official Responses and Stakeholder Positions

As the debate over global AI governance intensifies, ideological fault lines have emerged between governments, civil society, and corporate stakeholders.

The Developing World Perspective

Global South leaders argue that current international AI discussions are dominated by Western powers and corporate lobbyists, largely ignoring issues of data sovereignty, fair labor practices in data labeling, and equitable access to compute infrastructure. Developing nations demand that any international body must prioritize capacity-building and technology transfer rather than acting merely as a Western-led policing mechanism.

The Tech Sector Stance

Major technology companies have expressed conditional support for "harmonized regulation" to avoid a fragmented patchwork of laws. However, industry lobbyists consistently push back against open-source restrictions, mandatory pre-deployment algorithmic audits, and strict liability rules for model outputs. They frequently argue that overly heavy-handed oversight will stifle innovation and cede technological leadership to geopolitical rivals.

Multilateral Organizations

The United Nations, UNESCO (where Gabriela Ramos serves as Assistant Director-General for Social and Human Sciences), and the OECD have increasingly called for normative guardrails. UNESCO’s Recommendation on the Ethics of Artificial Intelligence, adopted by 193 member states, provides an ethical compass, but lacks enforcement teeth. UN leadership has repeatedly warned that voluntary corporate commitments are insufficient to mitigate systemic risks.

Implications: The Path Forward

The failure to establish a coordinated international framework for artificial intelligence carries profound, long-term implications for global stability, economic equity, and democratic integrity.

1. The Proliferation of "Wild West" Technologies: Without binding global standards, rogue actors, state-sponsored cyber-warfare units, and unscrupulous corporations will face few deterrents against deploying dangerous, unverified algorithms.
2. Erosion of Democratic Sovereignty: When multi-trillion-dollar corporations wield more influence over information ecosystems and critical infrastructure than elected governments, the very definition of national sovereignty is called into question.
3. The Imperative for a New Multilateralism: Just as the post-World War II international order established institutions like the International Atomic Energy Agency (IAEA) to manage nuclear proliferation, the modern era demands an equivalent institution for artificial intelligence—an "International AI Agency."

What an International AI Body Must Do

To be effective, this proposed international entity must possess a specific, robust mandate:

• Establish Global Safety Baselines: Set mandatory technical thresholds for frontier models regarding safety, transparency, and alignment.
• Independent Auditing: Maintain an independent pool of researchers and engineers capable of auditing proprietary models before public deployment.
• Resource Redistribution: Facilitate compute-sharing initiatives and AI capacity-building programs to ensure developing nations can harness AI for public good (such as climate adaptation and healthcare).
• Enforceable Accountability: Provide a centralized mechanism for dispute resolution and sanctioning when corporate or state actors violate established global safety protocols.

The window to establish orderly, cooperative governance is closing rapidly. As AI systems grow more autonomous and economically pervasive, leaving their trajectory to market forces alone is a catastrophic gamble. Governments must transcend short-term geopolitical rivalries and build the institutional architecture required to ensure that artificial intelligence serves humanity as a whole, rather than a privileged few.