Proposal for an International AI Governance Organisation to Prevent Catastrophic AI Outcomes

Bayer Judit (2026) Proposal for an International AI Governance Organisation to Prevent Catastrophic AI Outcomes. In: Szemelvények a BGE kutatásaiból IV. kötet. Budapesti Gazdaságtudományi Egyetem, Budapest, Magyarország, pp. 1-8. ISBN 978-615-6886-28-6

[thumbnail of BGE Szemelvények 2025_3.cikk.pdf]

Szöveg
BGE Szemelvények 2025_3.cikk.pdf - Megjelent verzió
Download (262kB)

Abstract

In light of escalating geopolitical tension and accelerating artificial intelligence (AI) developments, this article will argue for the formation of an International AI Governance Organisation spearheaded by the European Union. The organisation’s core mission will be the prevention of catastrophic outcomes from Artificial General Intelligence (AGI) or autonomous AI as well as rogue use of frontier AI systems including existential threats to humanity. The EU will be suggested as a natural leader given its strong regulatory legacy and global credibility. Positioning AI as a shared security risk – not merely a matter of innovation or competitiveness – the paper will argue for a narrow mandate focused solely on frontier risks. This mandate will aim to foster global consensus, and transcending value-laden ethical conflicts. Drawing on comparative insights from organisations such as the Internet Corporation for Assigned Names and Numbers (ICANN), the Institute of Electrical and Electronics Engineers (IEEE), the International Atomic Energy Agency (IAEA) and the World Trade Organisation (WTO), the article will lay out the institutional design of the proposed International AI Governance Organisation as a multistakeholder body comprising states, industry, researchers, and civil society. Equal representation will balance geopolitical power with technical and public interest input. Crucially, the paper will introduce application-programme-interface-level governance as a novel, artificial bottleneck and enforceable policy lever. In addition, the study will argue that transparency, traceability, and ethical use of frontier AI systems can be achieved via controlled application-programme-interface (API) access. In parallel, the article will explore the feasibility of creating a new ICANN-backed top-level domain (.aiapi) for licenced APIs, mirroring domain name system governance mechanisms to enhance oversight. The article will discuss that this approach could sidestep traditional enforcement weaknesses demonstrated by international law through leveraging functional chokepoints rather than sovereign coercion. It will also identify potential risks – especially organisational capture – and offer mitigation strategies through institutional checks, transparency, and civil society oversight. Finally, the study will position the proposed organisation not only as an urgent necessity but also as a strategic opportunity for the EU to exercise global leadership in shaping a cooperative AI future.