American policymakers are laser-focused on the need to lead in global AI markets and achieve advanced frontier capabilities before China. In pursuit of this goal, American policymakers have pursued a “full stack” export strategy, which promotes global adoption of vertically integrated American AI infrastructure, from hardware to models to cloud services. China is likewise seeking to build out global usage of its own AI models, though with a different approach — rather than export full-stack frontier model access, China is leading in open-weight model releases and supporting low-cost adoption of its AI, particularly in the Global South. This geopolitical contest has significant implications for the rest of the world, which is drawn into the push-and-pull between China and the U.S. as each seeks to build up global user bases for its own AI offerings.

Openness and AI for Good

But at the AI for Good Summit in Geneva, geopolitics were not the focal point. Instead, the Summit eschewed the geopolitics-competition-security nexus for discussions of openness, accessibility, and global equity. Presentations focused on AI’s potential for democratic resilience, energy efficiency, education, medicine, and more. Booths around the Expo showcased modular, privacy-enhancing edge AI that can run on your iPhone; robot pets to combat loneliness; assistive mobility devices; and a multitude of other AI applications intended to benefit the common good. Policy discussions revolved around increasing access to AI in the Global South and utilizing technology to boost growth and development around the world.

Notably missing from the summit was meaningful discussion of how choosing a side in the geopolitical race could affect these goals. This is partly consistent with the Summit’s orientation toward “AI for Good,” which prioritizes social impact over strategic competition, but even in discussions of access and deployment, many speakers implicitly or explicitly distanced themselves from great power dynamics. Instead, panels often focused on “AI sovereignty,” or the idea that countries should control their own development and deployment of AI and reject dependency on foreign AI providers in order to best harness AI for the common good. The chief pathway to sovereignty mentioned at the Summit was openness — open-source and open-weight AI data, software, and hardware, which can be adopted for free or for low cost, modified to fit specific use cases in the Global South, and run locally or on mobile devices.

The push for AI sovereignty is becoming increasingly popular around the world. Global South countries, for example, are eager to reap the benefits of AI but are particularly concerned that being forced to pick a side in the US-China AI race (by relying on one tech stack or the other) may increase foreign dependencies and limit opportunities for growth and development in the long term. Europe is also adopting a sovereign AI approach, as outlined in the European Commission’s June 2 Tech Sovereignty Package, which is intended to spur European advancements in open-source, hardware, and AI infrastructure and reduce reliance on foreign AI.

The Cost of Sitting Out

The world’s desire for AI sovereignty has deep implications for American AI strategy. The U.S. has made some limited efforts to appeal to markets desiring AI sovereignty — at the Summit, the U.S. booth featured representatives from webAI, a company offering a “full stack platform for sovereign AI.” The platform offers edge AI that runs securely on-device, and it specifically targeted sovereign AI buyers — though still via a full-stack approach. But China is still very much the leader in the global sovereign AI race.

China is outspokenly supporting openness, sovereignty, and the ideals espoused by many at the AI for Good Summit. Moreover, China leads the world in producing open-source and open-weight AI models, and its rhetoric on AI has centered accessibility, modularity, and cost-effectiveness in the Global South. At the Shanghai AI Summit on July 17, President Xi delivered a speech promising to help the Global South “bridge the AI and digital divides, promote sustainable development, and prevent creating new historical injustice in AI” — not so subtly echoing fears from the Global South that dependence on American AI will deepen global inequalities.

China is also lending political support to the Global South’s AI interests: On July 16 at the Shanghai AI Summit, China announced the establishment of the World AI Cooperation Organization (WAICO), a new intergovernmental AI governance body comprising 29 countries and whose mandate concerns upholding the UN Charter and a safe, ethical, and people-centric approach to AI. Member states are mainly from the Global South, and no G7 members are represented in the group. UN Secretary General Antonio Guterres was present at the signing of the agreement, emphasizing China’s recent alignment with the UN’s sovereignty- and openness- orientation towards AI policy, which was so evident at the AI for Good Summit the week before.

In other words, Chinese AI offerings — at least at face value — appear to be more conducive to global AI sovereignty goals than those of the U.S., and China is leaning into that dynamic. This impression is only strengthened by the White House’s recent blockage of access to Anthropic’s advanced Mythos model for foreign nationals (including Anthropic’s own employees), which sends an alarming signal to the rest of the world about the stability of reliance on imported American AI infrastructure.

All of that said, Chinese models themselves are not necessarily the draw here — the promise of sovereignty is. The Summit suggested that global demand for AI does not currently align with a binary choice between competing national ecosystems, and that many stakeholders prefer models that emphasize flexibility, local control, and shared access instead of full-stack imports.

If unaddressed, the gap between the AI the world wants and the AI the U.S. is providing will shape how U.S. AI initiatives are received abroad in contexts where openness, affordability, and sovereignty are paramount concerns — and China will continue to entrench itself as the leader in sovereign AI applications. American policymakers ought to take this misalignment seriously. If the U.S. wants to lead the world in AI, it needs to better understand and tailor offerings to the AI preferences of the rest of the world.