As New York City hosted global leaders for the United Nations General Assembly last week, Washington was consumed with Chinese President Xi Jinping’s state visit on Sept. 24. Artificial intelligence loomed large in both cities, with New York the epicenter of AI talks involving a wide range of global leaders, while Washington narrows its geopolitical lens to one bilateral relationship. The dichotomy is stark and raises a question: How would specific bilateral commitments between China and the United States impact each nation’s broader, existing foreign affairs strategy for AI?
Washington and Beijing have developed clear—and sharply divergent—diplomatic strategies to influence how other nations will develop, deploy, and govern AI systems and establish global AI leadership in the coming decades. Washington has emphasized strategic competition and American market dominance, bundling access to advanced chips, cloud services, AI models, compute power, applications, and financing into a “full-stack” AI export package that features U.S. industry and would be available to favored nations.
As New York City hosted global leaders for the United Nations General Assembly last week, Washington was consumed with Chinese President Xi Jinping’s state visit on Sept. 24. Artificial intelligence loomed large in both cities, with New York the epicenter of AI talks involving a wide range of global leaders, while Washington narrows its geopolitical lens to one bilateral relationship. The dichotomy is stark and raises a question: How would specific bilateral commitments between China and the United States impact each nation’s broader, existing foreign affairs strategy for AI?
Washington and Beijing have developed clear—and sharply divergent—diplomatic strategies to influence how other nations will develop, deploy, and govern AI systems and establish global AI leadership in the coming decades. Washington has emphasized strategic competition and American market dominance, bundling access to advanced chips, cloud services, AI models, compute power, applications, and financing into a “full-stack” AI export package that features U.S. industry and would be available to favored nations.
Beijing has promoted a vision for widespread collaboration based in AI as a development priority, with a focus on offering AI models that can be locally owned and operated, adapted for local needs, and supported by access to training and capacity building, a new multilateral AI body, and Chinese applications. Washington’s strategic partnerships have been designed to position the United States as the undisputed winner of an AI race with China. China’s partnerships have been designed to emphasize multilateral inclusion and equitable access to the benefits of AI, particularly for middle- and low-income countries.
China has been taking steps to move its foreign policy from rhetoric to practice, even though announcements still outpace implementation. On the eve of China’s World Artificial Intelligence Conference (WAIC) in July, representatives from 29 countries signed an agreement establishing the World AI Cooperation Organization (WAICO) as an “independent” intergovernmental organization headquartered in Shanghai. (Additional countries are now reported to have signed on, according to Chinese media outlets.) The following day, Chinese President Xi Jinping pledged 5,000 AI training and seminar opportunities for developing countries over five years. He also announced plans for international AI application centers with leading regional organizations such as BRICS, the Association of Southeast Asian Nations (ASEAN), the African Union, and the Arab League, as well as assistance for 30 countries to use China’s AI-powered Mazu meteorological early-warning system.
It remains unclear exactly how WAICO’s commitments would play out in practice. The July announcement described its status and aims but did not clarify a financing mechanism, offer a delivery structure for its commitments, or clarify how the organization’s independence would be protected and assured. It is clear that WAICO is designed to help China realize its stated AI strategy. For example, China’s July AI cooperation and development action plan called for accessible computing services for developing countries and support for adapting open AI models to local needs, a sentiment Xi echoed in his speech at the WAIC. (Open-weight models allow developers a model they can copy and adapt for their own purposes, without requiring them to train a model from scratch. This can make AI systems cheaper and more affordable and can be particularly useful for nations that need to build systems to operate across highly specific, local environments.)
Meanwhile, China’s existing cooperation with ASEAN may illustrate how regional partnerships would operate in practice: The ASEAN-China Artificial Intelligence Industry Innovation Center, inaugurated in Beijing on May 24, forms part of the two sides’ digital cooperation agenda. Its planned work pairs practical demonstrations of AI applications in sectors such as healthcare and agriculture with training for ASEAN officials, company employees, and students, connecting access to AI technologies with the skills needed to use them. With more than 3,000 exhibits from over 1,100 companies, the WAIC displayed the industrial breadth that could underpin Chinese AI applications.
Washington has begun implementing its own competing multilateral AI offer through Pax Silica, which seeks to secure the supply chains needed to build and power AI systems and has garnered support from approximately two dozen U.S. partners and allies. Efforts like the American AI Exports Program support U.S. consortia selling systems abroad, giving members priority access to federal financing and export-license processing. The departments of commerce and state have announced a range of new programs aimed at strengthening ties between American technology companies and foreign markets, including mechanisms to connect U.S. suppliers to foreign buyers, credential AI supply chains, and finance commerce-designated export packages.
While the U.S. approach offers components of an AI ecosystem strategy, no U.S. program creates a soft-power offering similar to China’s development offerings. The U.S. has also expressly rejected proposals for any form of independent multilateral body that would support broader AI governance or cooperation.
After the WAICO announcement, Washington indicated that it might even go a step further toward isolationism by positioning Pax Silica as an all-or-nothing proposition for its signatories. Reuters reported a draft letter from the State Department announcing that membership in Pax Silica “[could not] be held alongside membership in duplicative initiatives whose expectations conflict with our own.” While Reuters could not establish if or when such a letter would be sent, it did quote an unnamed U.S. official who said the letter was drafted to clarify that governments “can’t have it both ways.”
China, meanwhile, has publicly defended countries’ right to choose AI partners according to their development needs, allowing Beijing to present its assistance as compatible with keeping other partnerships open. That stance aligns more closely with what other governments are seeking when it comes to AI. On Sept. 21, Finland and Norway, along with 19 other nations and the European Commission, issued a joint statement urging global collaboration to address emerging frontier AI risks. The signatories included members of Pax Silica, WAICO participants, and nonaligned states.
Nations are not only aligned in seeking more multilateral mechanisms; they are also aligned in their desire to avoid dependency on any individual nation’s technology stack. On Sept. 14, European Central Bank President Christine Lagarde warned that reliance on foreign AI raises questions about whose laws govern sensitive data and whether access to essential services could be withdrawn or used as leverage in political negotiations. Governments weighing those risks lose nothing by increasing collaboration with Beijing, and pressure from Washington to forgo alternative partnerships and deepen dependencies on the U.S. as a technology partner seems increasingly foolhardy in a geopolitical environment where partners’ and allies’ trust in the U.S. has hit all-time lows.
It is hard to see how the U.S. can “win” any AI race by forcing governments around the world to choose between the two leading powers. Governments need support that fits the systems they already use and leaves them room to adopt AI on their own terms. They will want, and need, to build systems that incorporate both American and Chinese AI technologies as they seek to harness the benefits of AI transformation. They may also look for broader support than the limited offerings that the U.S. currently provides.
WAICO gives Beijing a standing organization through which it can organize this cooperation. Its appeal to partner governments is the prospect of practical help as their needs change, with membership in an institution devoted to AI development. A government could begin with one public service application and return for support adapting it to another language or extending it to other agencies. Whether WAICO can provide that assistance, and how much influence members will exercise over its priorities, remains unclear.
In addition, governments seeking to import and scale AI systems need to know whether a system will work with their data, meet local requirements, and remain affordable to operate—requirements that China has named explicitly in its technology diplomacy strategy. AI export packages that meet those needs early in adoption will secure a critical first-mover advantage for the nation (or companies) behind them.
Training programs will shape local engineers’ familiarity with specific software or hardware tools, and procurement rules built to support early purchases will determine which technologies can be deployed, how, and by whom.
First-mover advantages won’t go to the AI leader with the most cutting-edge models. It will go to the country that can best support partner nations in connecting infrastructure, models, applications, financing, capacity building, and operational standards into a coherent offering that allows leaders to scale AI adoption while maintaining their right to mix and match systems as needed.
The world is waiting to see what, if any, agreements the U.S. and China might strike on managing AI risks. For the United States, any agreement undercuts its ability to demand that partners and allies refrain from collaboration or trade with China regarding their own AI needs. For China, a stated commitment to openness and inclusion will be challenged by a singular agreement with the United States that excludes all other nations from access to critical information sharing on AI risks and safety standards. If the U.S. and China can agree to cooperate on critical AI safety risks, world leaders will breathe a collective sigh of relief. And then, with their next breath, they will ask, “Where do we come in, and what does this mean for us?”
At the moment, Beijing is far better positioned to answer that question than Washington.
