China built an AI empire for the global south. Trump has a better offer

China put “World” in the name of its new artificial intelligence organization. It left out the United States. On July 16, 29 countries signed an agreement in Shanghai establishing the World Artificial Intelligence Cooperation Organization, with its headquarters in China.Less than seven weeks later, President Donald Trump’s administration supplied the beginnings of an answer. On Sept. 2, innovation ministers from the Group of 20 — including representatives from China, the European Union, India, Japan, South Korea, Britain, and other major economies — issued a consensus statement on emerging technologies. They agreed on pro-innovation policy, workforce development, intellectual property, standards, and investment in the industrial supply chains that make AI possible.That contrast matters more than another contest between diplomatic acronyms. Beijing created a new institution bearing a global name. The Trump administration used an existing forum and obtained agreement from countries across the emerging technological divide. One announced a headquarters. The other demonstrated convening power.

WAICO should not be laughed away. The broader Chinese offer answers genuine needs. Beijing has pledged 5,000 AI training opportunities for developing countries, cooperation centers with organizations including the African Union and the Association of Southeast Asian Nations, and access to its AI-enabled Mazu meteorological warning system for 30 countries. For governments lacking advanced computing infrastructure, technical personnel, and affordable models, those are meaningful offers.Washington also cannot dismiss China as a permanent technological runner-up. The 2026 Stanford AI Index reports that the performance gap between the best American and Chinese models has effectively closed, with the leading U.S. model ahead by 2.7% as of March. Chinese developers have also built a powerful open-weight ecosystem: users can download many model weights, adapt them, and run them outside the original company’s application programming interface. A Stanford study of that ecosystem found Chinese model families spreading rapidly across global developer platforms.That technological credibility makes WAICO serious. It does not make it global.WAICO can survive as an organization. It can train officials, coordinate projects, shape technical standards, and give participating governments a louder voice. It could become an influential institution across the global south. None of those outcomes requires American membership.WAICO’s advertised ambition is larger: to help shape global AI governance. Global rules must operate across the world’s principal model, chip, cloud, research, and investment ecosystems. The U.S. and several other leading AI economies remain outside the founding group. A safety standard observed inside one technological sphere remains a bloc standard. A global rule that one of the two leading AI powers will not implement is global in name alone.The same is true in reverse. An American-led AI institution without China could coordinate allies and protect trusted supply chains, but it could not credibly claim to govern AI everywhere. The technology crosses borders even when institutions do not. Any genuinely global regime will eventually need both Washington and Beijing, however uncomfortable that fact may be.China’s strongest sales pitch is the apparent contrast between Chinese openness and American control. Leading American frontier companies generally deliver their most capable models through proprietary products, cloud services, and APIs. The provider retains the weights, controls updates, applies safeguards, and can change the terms of access. Many Chinese models, by contrast, are available in open-weight form. A government or company with sufficient computing resources can download them, customize them, keep sensitive data locally, and continue operating if relations with the original provider deteriorate.For a developing country, the distinction can be decisive. An American API may provide outstanding performance, but it remains permission to use someone else’s system. Downloaded weights look more like possession. A somewhat less capable model that a government controls locally may be more attractive than one that can be repriced, restricted, or withdrawn from abroad.Yet “America closed, China open” is a commercial snapshot, not a civilizational truth. Open-weight does not necessarily mean open-source: training data, training code, safety testing, and development decisions may remain hidden. Beijing also subjects domestic generative AI services to regulation. The Trump administration has not conceded openness. America’s AI Action Plan describes open-source and open-weight models as having geostrategic value and calls for a supportive American open-model ecosystem.China does not own openness, and America should not let Beijing market it as a Chinese invention.This week’s G20 result shows the right American response. The Carolina Principles for Emerging Technologies favor flexible policy, reliance on existing sector-specific regulation where it works, and new rules limited to genuinely novel gaps. The ministerial statement also preserves national sovereignty while supporting international engagement. Its AI Prosperity initiatives emphasize technical training and private-sector partnership. China joined the consensus. That is a more consequential answer to WAICO than joining Beijing’s institution or pretending it does not exist.Trump should resist answering a Chinese bureaucracy with a larger American bureaucracy. Washington already possesses stronger instruments: innovative companies, allied markets, research universities, capital, advanced chips, cloud infrastructure, and the new G20 initiatives. In August, the State Department also opened the Pax Silica AI Assistance Project to build trusted supply chain infrastructure in partner countries. The task is to combine those advantages into an offer that countries can actually use.First, the administration should give the AI Prosperity Compact tangible content: affordable computing access, technical training, local-language development, and public-private deployment projects. China is gaining influence by solving capacity problems, not merely by publishing communiques. America must compete at the same practical level.Second, U.S. providers should offer partners greater continuity and portability. Clear access conditions, reasonable transition periods, exportable operational records, and open-weight or locally deployable options, where security permits, would reduce the fear that national capabilities rest on a revocable foreign subscription. Countries should not have to choose between American quality and operational sovereignty.Third, Washington should not treat every government accepting Chinese training or infrastructure as politically lost. Telling developing countries to reject useful technology is not a strategy. Giving them a more capable, transparent, secure, and dependable alternative is.WAICO does not need the U.S. to hold meetings, open offices, or influence its members. It may do all three successfully. But a China-backed body launched without America cannot become the world’s AI authority by placing “World” in its title. That status must be earned through broad participation, technical interoperability, credible verification, and rules that the leading AI powers will actually follow.OPINION: LEAVE YOUR AI PANIC NEXT TO YESTERDAY’S NUCLEAR POWER OPPOSITIONChina has built a secretariat. The Trump administration has assembled a pro-innovation consensus broad enough to include allies, emerging powers, and China itself. Now it must attach access, infrastructure, and reliability to that consensus.If it does, Trump will not need to defeat WAICO or seek a seat inside it. He will have answered China’s claim to the word “World” with an offer the world can actually use.Burak Oktenli is an independent researcher based in Washington, D.C. He holds a bachelor’s degree in computer science and engineering from the University of South Florida and a master’s of business administration, and is completing a master’s of professional studies in applied intelligence at Georgetown University. His writing focuses on defense strategy, alliance burden-sharing, military technology, and the governance of autonomous systems.