In the last week of September 2026, six Asia-Pacific governments used the general debate of the 81st UN General Assembly to say what they want from artificial intelligence. The capabilities are moving faster than any shared rule for them. These speeches are the public record of how the region’s governments would meet that gap: who stays accountable, who the technology is for, who may build it on their own terms, and who is in the room when the rule is written.

Singapore: a shared rule

Singapore called AI “arguably the most urgent frontier.” Foreign minister Vivian Balakrishnan set out three risks. Autonomous systems can slip out of control. Rogue actors could use the tools to make biological weapons, other weapons of mass destruction, or weapons of mass disruption. Economic, social and political disruption is, he said, only just emerging.

He argued that a halt is already too late. Rivalry among the largest powers, and the money behind frontier models, will keep the work moving. What he asked for instead was pace with limits: testing and evaluation before deployment, clear limits on what an autonomous system may do, and a way to intervene when a system behaves in a way people did not intend. “Human beings must ultimately remain in control and accountable.” The picture he left in the hall was a button. Do you want the nuclear button in the hands of a person, or in the hands of a model?

On what that would mean in practice, the speech is specific but limited. Singapore said the UN should explore all options, “including the idea of an UN Framework Convention on AI Safeguards.” Balakrishnan also said an international body might one day do for AI some of what the International Telecommunication Union does for technical standards, or the International Atomic Energy Agency does for verification.

South Korea and China: who gets the benefit

On 22 September 2026, South Korea's President Lee Jae Myung put a different offer first among three visions he called a K-Initiative. The name in the UN text is a “Global AI Universal Basic Society.” No country and no individual, he said, should be left out of the basic benefits of AI development.

South Korea tied that to rights it already lists in social policy: an “AI Universal Basic Health” programme, then social security, labour, housing and culture. It also said AI should become a tool for climate, public health, food insecurity and refugees. The anxiety of young people about jobs is in the speech. The method he offered the hall is a “Global AI Hub,” a platform with international organisations, development banks, business and academia.

China’s line, four days later, sits beside Korea’s and is shorter. Vice President Han Zheng, in the English text published by the Ministry of Foreign Affairs, said governments should take a people-centred approach, develop AI “for the positive and for good,” narrow the digital divide, and promote development that is universally beneficial. The paragraph sits inside a wider argument about development.

South Korea is offering a hub and a benefits floor. China is offering a principle about who the technology is for.

India: a model built around national priorities

India’s national statement on 26 September, delivered by External Affairs Minister S. Jaishankar, contains two lines about AI, and they do not point the same way.

In the passage on the Global South, he says the march of technology has become a source of anxiety, and that misuse of AI is “sharpening bias, and diluting mutual respect.” Later, AI is “a transformational driver” for human empowerment, economic prosperity and population-scale problem solving. India hosted an AI Impact Summit in February 2026. What he then states as a right, and what separates this speech from Singapore’s shared rule, is that each nation may shape its own ecosystem, “with the development of its own models, and deployment strategies, consistent with its national goals and priorities.” AI, he said, will be part of India’s partnership agenda with the Global South.

Korea’s speech asks that the basic benefits reach every country and person. India’s speech keeps the right to build the system around national priorities. Jaishankar states both the warning and the right. He does not choose between them.

Japan: a code and another summit

Japan spoke on 22 September, the same day as Korea. Prime Minister Sanae Takaichi’s address treats the Hiroshima AI Process as work already done. In the speech, Japan launched that process and established what it calls the world’s first international principles and code of conduct for generative AI. That “world’s first” is Japan’s description.

Japan's proposal sits somewhere short of a treaty. It builds on the Hiroshima AI Process. Japan said it is working with like-minded partners and Global South partners on a “safe, secure and trustworthy” AI environment, and that it will increase cooperation on the risks of highly capable frontier systems. To accelerate that, Japan intends to host an AI summit “at the earliest opportunity.”

Set next to Singapore’s idea of a convention, Japan is pointing at a code it says already exists, and at a meeting it intends to convene.

Malaysia: a seat in the decision

Mohamad Hasan, Malaysia’s foreign minister, on 26 September, asked a prior question: who governs the transformation. He said a handful of private actors now control how billions of people communicate, work, transact and consume information. Citizens, he said, are seeing their governments choose not to regulate AI, and the ways it is being abused, and instead give it more power. The opportunities are enormous, and so are the risks. Technology, he said, must serve humanity. The answer cannot be that a small number of companies or countries decide the rules for everyone else.

The most distinctive line in Malaysia's statement was this: “The Global South cannot simply become a destination for data centres, while decisions about A.I. are made elsewhere.” Partners in innovation, he said, not merely customers and markets, and not merely destinations for investment. He also said the water, energy and land that the infrastructure consumes have to be managed if the benefits are to be shared across borders.

Malaysia is asking to be in the room when the rule is written.

The six speeches leave a clear record of what these governments want from AI. Singapore wants safeguards, and a person who remains accountable, and it has asked the UN to explore a framework convention. Korea wants the basic benefits to reach every country and every person, and it has named a hub to organise that. China, more briefly, wants the technology developed for people and shared widely. India wants each country free to build its own models around its own priorities, even as it warns that misuse is sharpening bias. Japan wants the Hiroshima code treated as work already done, and a summit to carry the cooperation further. Malaysia wants a say in the rule for countries that would otherwise host the infrastructure while the decisions are made somewhere else.

What to watch

• Whether any government tables a text for a UN framework convention on AI safeguards, or leaves Singapore’s proposal at the level of an idea.

• Whether Korea’s Global AI Hub acquires members, a host or a budget, or remains a name in a speech.

• Whether Japan gives its intended AI summit a date.

• Whether India’s right to “its own models” shows up in a procurement rule, a funded programme, or its Global South partnership.

• Whether Malaysia’s objection, decisions made elsewhere, is taken up by any other government as a drafting demand.