AI Faces Global Governance Push
As researchers convened at Dartmouth to mark 70 years since the term "artificial intelligence" was coined, the United Nations established a new international scientific panel to guide global regulation. The move comes as the 2025 Stanford Index highlights both remarkable performance gains in AI and a growing gap between top systems and public accessibility. The UN panel will assess AI's risks and opportunities to provide evidence-based recommendations.
The original 1956 Dartmouth Summer Research Project on Artificial Intelligence was an ambitious two-month, 10-man study funded by a modest $7,500 grant from the Rockefeller Foundation. The project's proposal, co-authored by John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon, put forth the foundational conjecture that "every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it." The UN's High-level Advisory Body on AI is composed of 39 experts from 33 countries, including executives from Microsoft, OpenAI, and Google-Alphabet, as well as academics and government officials. Their final report, "Governing AI for Humanity," was the result of extensive consultations involving over 2,000 participants worldwide. This global governance push is happening alongside other significant international efforts. The Council of Europe has established the first legally binding international treaty on AI, and other initiatives include the G7's Hiroshima AI Process and the OECD AI Principles. These agreements aim to create interoperable policies as AI development and its supply chains are inherently global. The 2025 AI Index shows the performance gap between top-tier AI models is shrinking rapidly. The score difference between the number one and the tenth-ranked model fell from 11.9% to 5.4% in just one year, with the top two models now separated by only 0.7%. While industry labs produced nearly 90% of notable AI models in 2024, open-weight models are increasingly closing the performance gap with proprietary systems. This trend, combined with rapidly falling costs—the price of running inference on open models is 87% less—is making advanced AI more accessible. The economic stakes are substantial, with private AI investment in the United States reaching $109.1 billion in 2024, over 12 times more than China's $9.3 billion. Meanwhile, enterprise adoption of AI has surged, with 78% of organizations reporting its use in 2024, a significant increase from 55% the previous year.
