Global Governance of Artificial Intelligence (AI)

Global governance of AI means the rules, agreements and institutions that countries are building together to manage the risks and benefits of artificial intelligence, since AI systems built or used in one country can easily affect people in other countries. Because no single country can control how AI is developed and used everywhere, governments are creating both national laws and international cooperation mechanisms side by side.

Why does it exist?

AI systems, especially powerful ones like large language models and autonomous "AI agents", can cause harm that crosses borders: a data breach, a biased decision-making system, or a malfunctioning autonomous system does not stay confined to one country. At the same time, AI offers major benefits like faster scientific research, healthcare improvements and economic growth. Global governance tries to allow these benefits while reducing risks like job losses, misuse for cyberattacks, disinformation, or loss of human control over powerful systems.

Where did it come from?

International AI governance is still a very young field. Key milestones include the UK's AI Safety Summit at Bletchley Park in November 2023, where 28 countries and the European Union signed the "Bletchley Declaration", agreeing to cooperate on AI safety. This was followed by the AI Seoul Summit in May 2024, the AI Action Summit in Paris in February 2025 (co-chaired by France and India), and the India AI Impact Summit in New Delhi in February 2026.

In parallel, the United Nations adopted resolutions on AI, and multiple countries and blocs, including the European Union with its AI Act, began framing their own domestic AI laws. The Global Partnership on AI (GPAI), launched in 2020, is another multilateral effort focused on responsible AI development.

What has the UN set up?

In September 2024, UN member countries adopted the Global Digital Compact, which promised two new UN bodies for AI. On 26 August 2025, the UN General Assembly created both through resolution A/RES/79/325, adopted by consensus (without a vote):

• Independent International Scientific Panel on AI: 40 experts from different regions, serving in their personal capacity for three years. The General Assembly appointed the first 40 members on 12 February 2026, by a vote of 117 to 2. Their term runs until February 2029. Each year it writes a report on the chances, risks and effects of AI, outside the military area. Think of it like the IPCC (the UN's climate science panel), but for AI.
• Global Dialogue on AI Governance: a meeting where every UN member country can discuss AI rules together. Its first session was held in Geneva in July 2026, and the next is due in New York in 2027. Its results are summaries, not binding rules.

How does it work?

Global AI governance currently works through a mix of mechanisms rather than one single treaty:

• Summits and declarations: high-level meetings where countries agree on shared principles, such as testing powerful AI models for safety before wide release.
• UN processes: the Scientific Panel on AI gives all countries the same scientific facts, and the Global Dialogue on AI Governance gives every country a seat at the table.
• Bilateral communication channels: direct government-to-government arrangements, like the one between the US and China, to share information quickly if an AI-related incident occurs, similar in spirit to existing hotlines used to manage other kinds of crises.
• National regulations with global reach: laws like the EU's AI Act, which apply to any company selling AI products into that market, effectively setting a global standard because companies often follow the strictest major market's rules everywhere ("the Brussels effect").
• Voluntary commitments by AI companies: major AI developers have made voluntary safety commitments, such as testing models before release and sharing safety research, though these are not legally binding.

The key parts and facts

The Bletchley Declaration (2023) is widely seen as the first major international agreement recognising the risks of "frontier AI" (the most advanced AI systems). The EU AI Act, adopted in 2024, is the first comprehensive AI-specific law by a major economic bloc, and classifies AI systems by risk level (unacceptable, high, limited and minimal risk), banning the highest-risk uses outright.

India does not yet have a dedicated AI law, but has released policy frameworks (such as through the Ministry of Electronics and Information Technology) promoting a "pro-innovation, risk-based" approach rather than a single comprehensive AI Act.

India's position and Indian examples

India has generally favoured a lighter-touch, sector-specific approach to AI regulation rather than one big AI law, relying on existing laws (like data protection rules under the Digital Personal Data Protection Act, 2023) alongside AI-specific guidelines. India also participates in international AI governance discussions, including the Global Partnership on AI, and has hosted its own AI summits and initiatives such as the "AI for All" strategy under NITI Aayog.

India co-chaired the Paris AI Action Summit (February 2025) with France. It then hosted the India AI Impact Summit at Bharat Mandapam, New Delhi, in February 2026, the first summit in this series held in the Global South. That summit adopted the non-binding New Delhi Declaration on AI Impact, guided by the idea "Sarvajan Hitaya, Sarvajan Sukhaya" (welfare for all, happiness for all).

As of late February 2026, 91 countries and international organisations had endorsed it. India's stand is that AI should be inclusive and serve the Global South.

Commonly confused concepts

• AI governance vs AI regulation: Governance is the broader set of norms, institutions and cooperative mechanisms (including voluntary and international ones); regulation refers specifically to binding legal rules enforced by a government, like the EU AI Act.
• Frontier AI vs narrow AI: Frontier AI refers to the most advanced, general-purpose AI models capable of a wide range of tasks and posing broader risks; narrow AI refers to systems built for a single, specific task (like a spam filter), which typically pose much smaller systemic risk.

Issues, criticism and the way forward

Critics argue that global AI governance is fragmented, moving more slowly than the technology itself, and dominated by a handful of countries (mainly the US, China and the EU) that host the largest AI companies, leaving developing countries with limited say. There are also concerns that voluntary commitments by AI companies are not enough, since they are not legally enforceable.

Experts and international bodies have called for a more coordinated global framework, possibly modelled on how countries cooperate on nuclear safety or climate change, so risks like AI-enabled disinformation, autonomous weapons and large-scale job displacement can be managed collectively rather than country by country.

Concepts to Know

• Frontier AI: The most advanced, general-purpose AI models available at a given time, capable of many different tasks and considered to carry the greatest potential risks and benefits.
• AI agent: An AI system that can take actions on its own, such as browsing the internet, using software tools, or accessing external systems, rather than only answering questions when asked.
• Global Digital Compact: A UN agreement adopted in September 2024 as part of the "Pact for the Future". It sets shared goals for the digital world, such as closing the digital divide and governing AI, and it led to the UN's Scientific Panel and Global Dialogue on AI.
• The Brussels effect: The tendency for a major market's regulations (often the European Union's) to become a global standard, because companies find it easier to follow one strict rulebook everywhere than to build different versions of their product for different markets.