The emergence of artificial intelligence (AI), catalyzed by the debut of technologies like ChatGPT approximately three years ago, has introduced a new, complex global challenge that is now at the forefront of discussions among world leaders and diplomats at the United Nations. While AI's capabilities have amazed globally, the rapid development by tech companies has also been accompanied by serious warnings from experts regarding its profound risks, including potential existential threats such as engineered pandemics and large-scale disinformation campaigns. These concerns underscore an urgent need for robust safeguards and international governance.

In a significant move to address these challenges, the United Nations has adopted a new governance architecture, marking its most substantial effort yet to establish control over AI development and deployment. This initiative follows previous multilateral attempts, including AI summits organized by Britain, South Korea, and France, which had only yielded non-binding pledges. A pivotal step was taken last month when the General Assembly passed a resolution to create two crucial bodies dedicated to AI governance: a global forum and an independent scientific panel of experts.

The global forum, officially named the Global Dialogue on AI Governance, is envisioned as a vital platform for governments and various stakeholders to engage in discussions concerning international cooperation, share innovative ideas, and develop practical solutions for AI. Its formal meetings are scheduled to take place in Geneva next year and in New York in 2027. Concurrently, efforts are underway to recruit 40 experts, including two co-chairs (one from a developed and one from a developing nation), for the independent scientific panel. This panel has drawn comparisons to the UN's influential climate change panel, suggesting its potential to guide global understanding and policy on AI.

Despite the symbolic triumph these new mechanisms represent, being hailed as "by far the world’s most globally inclusive approach to governing AI," skepticism remains regarding their practical efficacy. Isabella Wilkinson of Chatham House noted that these mechanisms might prove "mostly powerless," citing concerns about the UN's traditionally "lumbering administration" attempting to regulate a rapidly evolving technology like AI. The UN Security Council is also set to convene an open debate to explore how it can ensure the responsible application of AI, compliance with international law, and its potential role in supporting peace processes and conflict prevention.

Amidst these UN initiatives, a group of influential experts, including senior figures from major AI developers like OpenAI, Google’s DeepMind, and Anthropic, has issued a call for governments to agree upon "red lines" for AI. They advocate for these minimum guardrails to be implemented by the end of next year, specifically designed to avert the "most urgent and unacceptable risks." These experts propose an internationally binding agreement on AI, drawing parallels with historical treaties that banned nuclear testing and biological weapons or protected the high seas. Stuart Russell, an AI professor at the University of California, Berkeley, articulated a straightforward concept: developers should be required to prove safety as a condition for market access, a standard applied to industries like medicines and nuclear power stations. He further suggested that UN governance could mirror the International Civil Aviation Organization, which coordinates safety regulators globally. Additionally, Russell proposed that diplomats create a "framework convention" that offers flexibility, allowing it to be updated continually to reflect the latest advancements in AI, rather than adhering to rigid, set-in-stone rules.