22nd September 2026 – (Washington) Artificial intelligence has become a central dividing line in the strategic rivalry between the United States and China, yet the two powers are also edging toward cooperation to manage its risks. As President Donald Trump prepares to host President Xi Jinping in Washington on 24th September, officials from both sides met in New York on Sunday to discuss establishing a channel on AI safety, with Washington proposing a mechanism to notify each other of serious AI incidents. The move comes despite persistent disagreements over technology development, governance and access to cutting‑edge hardware.
The United States currently holds the lead in advanced AI models and high‑end chips, though China is narrowing the gap. Analysts note that companies in both countries dominate state‑of‑the‑art model development. Washington has sought to protect its advantage by restricting Chinese access to top‑tier semiconductors from American firms, prompting Chinese companies to lean into powerful open‑source systems from players such as DeepSeek and Moonshot AI while extracting greater efficiency from constrained hardware. Chinese developers have pushed rapid deployment of AI into products and services, whereas several major US labs are racing to build frontier models with stronger reasoning and organisational capabilities.
Rising model capability has amplified global concern about potential misuse. Leading developers, including Anthropic and OpenAI, have warned that more powerful systems could assist malicious cyber activity, enable biological threats or behave in unanticipated ways. This shared risk profile has created a narrow but significant area of common ground for Washington and Beijing: preventing the most capable AI tools from falling into the hands of rogue actors. Experts argue that neither side wants to see such capabilities exploited to harm international security.
Beyond bilateral talks, both governments are building competing international frameworks. The United States launched Pax Silica in December 2025 to bolster secure and trusted technology supply chains among partners, spanning critical minerals, semiconductors, advanced manufacturing, energy and digital infrastructure, with signatories including Japan and South Korea. China in July unveiled the World Artificial Intelligence Cooperation Organization, or WAICO, positioned as a platform for collaboration on AI development and governance, drawing founding members largely from the Global South and pitched as a route to help developing nations build capabilities.
Southeast Asia has emerged as a focal point of these efforts. Under Pax Silica, Washington introduced AI SPARK in July to provide financial and technical support for governments in the region adopting American AI for public‑sector use. Beijing is promoting a broader digital ecosystem and the uptake of lower‑cost open‑source models to expand access. Singapore and the Philippines have joined Pax Silica, while Indonesia, Malaysia, Cambodia, Laos and Myanmar are part of WAICO; Singapore has also been invited to join WAICO and is weighing the offer, with officials noting there is no Association of Southeast Asian Nations position on either initiative.
Policy specialists caution that both frameworks still lack detail. While headline goals are clear, practical commitments and obligations remain to be defined. Observers say Pax Silica is designed to harden AI supply‑chain resilience amid tightening Chinese export controls, yet participant requirements are still under discussion. Technically, countries can blend elements from both ecosystems, for example running Chinese open‑source models on servers powered by US chips, or pairing American hardware with components sourced through Chinese supply networks. The extent to which such “mix and match” strategies remain viable may ultimately hinge less on technology and more on geopolitics, including whether either side pressures partners to exclude rival systems.
Cost and control are paramount for developing economies. Many are prioritising models that are affordable to run at scale and can be embedded across public services and business operations. Open‑source systems are appealing because they are often free or lower‑cost and can give users greater control over data and availability; once downloaded, they cannot be switched off by a foreign provider. Yet both the US and Chinese initiatives are still nascent, and analysts argue that broader input from prospective members—especially in Asia and the Global South—will be crucial. Their success will depend on whether they can incorporate a diverse set of national interests and secure meaningful participation beyond their own strategic objectives.