Can the United States and China simultaneously cooperate and compete on artificial intelligence? This fundamental question currently drives intense debates in Washington regarding the wisdom of pursuing diplomatic talks on the subject with Beijing. Successive administrations have grappled with this dilemma: the Biden administration paired sweeping technology restrictions with a narrow, targeted dialogue on technological risk, while the Trump administration has kept communication channels open even as it has downplayed safety concerns. Some policy experts go even further, nursing the hope that cooperation on artificial intelligence safety can be siloed off from great-power competition entirely—a belief that is as politically appealing as it is fundamentally fanciful. In practice, any dialogue concerning artificial intelligence risk takes place entirely within the broader, unyielding strategic competition between the United States and China.
Beijing has long understood and appreciated this dynamic, whereas much of Washington has not. While the United States has shifted its approach to artificial intelligence competition and safety across and within presidential administrations, Beijing has kept its artificial intelligence diplomacy laser-focused on gaining a strategic competitive advantage over the United States. China has certainly taken part in international artificial intelligence safety discussions, but it has routinely used them—alongside bilateral talks with the United States—to paint American export controls as the primary obstacle to global technological cooperation. By doing so, China has successfully cast itself as the champion of the developing world, promising a Chinese-led artificial intelligence ecosystem free of technological or financial barriers as an appealing alternative to what it characterizes as a restrictive, U.S.-led global order.
This pitch is clearly working. A growing number of sovereign states are actively embracing China’s message, and Chinese open-weight models are rapidly gaining users across the developing world. If the United States wants to maintain its technological lead and make meaningful, verifiable progress on safety, it must fundamentally change how it operates. Washington needs to treat every diplomatic dialogue as an integral part of the overarching competition with China. To achieve this, the United States must ramp up strategic pressure on China outside of formal talks and offer nations still choosing sides a far more compelling vision for artificial intelligence adoption and governance. As U.S. and Chinese officials prepare to meet again in November, Washington should seize the opportunity to test Beijing’s true seriousness regarding safety. In short, the United States must approach artificial intelligence diplomacy with the exact same strategic clarity that China has displayed for years.
Chipping Away at the Global Order
The first round of advanced semiconductor export controls enacted by the Biden administration in October 2022, closely followed by the public release of ChatGPT just a month later, signaled to the world that the United States had taken an early, commanding lead in the global artificial intelligence race and would act decisively to weaken China’s artificial intelligence industry. Washington and its key allies also seized the initial diplomatic initiative, taking the first concrete steps toward establishing global artificial intelligence governance. In May 2023, the G-7 began intense discussions on artificial intelligence governance that would eventually produce a comprehensive code of conduct for software developers. Just weeks later, the United Kingdom announced that it would host the very first global artificial intelligence safety summit at Bletchley Park.
China responded swiftly by rolling out a meticulously crafted diplomatic script that it has continued to hone ever since. The core tenets of this strategy first appeared in the Global Artificial Intelligence Governance Initiative, published in October 2023. The initiative called for countries to "put equal emphasis on development and security"—conveniently leaving the definition of security entirely up to individual states—while sharply contrasting China’s focus on cooperation and development against U.S. export controls. It attacked Washington for "creating barriers and disrupting the global artificial intelligence supply chain through technological monopolies and unilateral coercive measures." Despite Beijing’s own rapidly growing body of domestic artificial intelligence regulations, the initiative offered little more than hollow slogans regarding the actual technical risks posed by the technology.
This basic formula—emphasizing cooperation and development while attacking U.S. restrictions, all while offering virtually nothing substantive to say about systemic risk—has guided China’s artificial intelligence diplomacy for the past three years. In March 2024, the UN General Assembly adopted a U.S.-led resolution on safe, secure, and trustworthy artificial intelligence intended to appeal broadly to both developed and developing countries. China initially supported the resolution, but quickly pivoted to introduce a rival resolution centered squarely on providing help for poorer countries and "emphasizing that member states should enjoy equal opportunities" to develop and use the technology. At a UN Security Council meeting on artificial intelligence chaired by the United States in December of that year, China’s UN ambassador, Fu Cong, once again counterposed Beijing’s stated interest in cooperation against "discriminatory barriers based on ideological differences." In thinly veiled terms, Fu charged that U.S. restrictions on China directly harmed the developing world.
To win the dedicated support of these nations as they built out their national artificial intelligence infrastructure, China recognized that it needed an institutional bloc of its own. In July 2025, Premier Li Qiang proposed the creation of the World Artificial Intelligence Cooperation Organization (WAICO), which promised developing countries advanced technology, specialized training, and a greater say in writing global artificial intelligence rules. In a speech marking its formal launch a year later, Chinese leader Xi Jinping called for "openness"—shorthand for guaranteed access to U.S. microchips and the widespread adoption of Chinese open-weight models. After briefly urging countries to take the risks of artificial intelligence seriously, Xi added that they should, "in the meantime," jointly oppose "overstretching the national security concept in the field of artificial intelligence and placing one country’s security over that of others." In Xi’s narrative, cooperation on safety risks was entirely subordinate to overcoming U.S. restrictions.
WAICO has quickly gained substantial standing among its target audience. An early endorsement by UN Secretary-General António Guterres lent the organization immediate international credibility in the eyes of developing countries, and Chinese officials have since promoted WAICO in frequent speeches and bilateral diplomatic meetings with countries as varied as Brazil, Pakistan, and Serbia. Beijing has heavily signaled that formal membership will come with tangible perks. Its draft export controls on leading models, released in October, explicitly promise to help WAICO’s 37 member states build their domestic artificial intelligence capabilities, suggesting that members could enjoy much easier access to models that remain heavily restricted for the rest of the world.
Xi personally led this high-stakes diplomatic offensive in the lead-up to the Washington summit with U.S. President Donald Trump in September. During a major speech in Kyrgyzstan, he pledged to expand a Chinese artificial intelligence center for the Shanghai Cooperation Organization designed to help member states build local language models and train local engineers. In Cairo, he and Egyptian President Abdel Fattah el-Sisi signed sweeping agreements covering artificial intelligence, data centers, and semiconductors, which have rapidly become a standard feature of Xi’s foreign trips. Furthermore, in New Delhi, Xi proposed an open-source artificial intelligence zone for BRICS members, telling assembled leaders that a "small yard, high fences" approach would only "box oneself in and divide the world." Notably, artificial intelligence risks were entirely absent from the agenda. The core message instead was that U.S. export restrictions would ultimately isolate Washington, not Beijing.
Much of the allure of China’s pitch rests heavily on how the country’s developers share their technology. Chinese developers routinely release their best models with open weights, meaning any foreign government or private company can download them, run them on its own hardware, and adapt them to its own native language. For governments naturally wary of depending entirely on American companies whose exports Washington can unilaterally restrict at any moment, open-weight models promise a far greater degree of national control. Chinese developers rarely conduct or publish the results of rigorous safety testing for these models, leaving external users to discover only after release whether the models can be exploited to launch sophisticated cyberattacks or build biological weapons. Yet, their sheer accessibility allows Beijing to argue persuasively to the developing world that any effort by the West to contain the risks of Chinese models is simply an attempt to deny artificial intelligence to poorer nations.
Changing the Subject in Bilateral Talks
China’s approach to bilateral artificial intelligence talks with the United States has deflected substantive discussions of risk just as assiduously as its global diplomacy has. In May 2024, the United States and China held formal talks in Geneva. The primary U.S. objective for these discussions was to test whether an in-depth, technical exchange with China regarding artificial intelligence risk was genuinely possible. The answer proved to be a definitive no. Chinese diplomats skillfully steered the discussion away from systemic risks for hours on end and declined to even explain China’s own domestic artificial intelligence regulations beyond a superficial, high-level overview, despite having technical experts from the relevant regulatory agencies present in the room.
The core of the Chinese position came through most clearly in the opening and closing remarks delivered by the delegation’s most senior officials. They began with a warm appeal for cooperation on capacity-building projects throughout the developing world. They then cast U.S. technology restrictions as a malicious attempt to contain China’s economic development, framing the U.S. emphasis on safety as an artificial barrier preventing developing countries from accessing and using modern technology. The strategic implication was unmistakable—American export controls were the primary impediment to any meaningful cooperation on artificial intelligence safety. The one concrete outcome of these diplomatic efforts—a November 2024 joint statement in which former President Joe Biden and Xi affirmed "the need to maintain human control over the decision to use nuclear weapons"—ultimately asked very little of Beijing in practice.
This September, following the summit between Trump and Xi in Washington, China’s Ministry of Foreign Affairs published an "eight-point consensus," representing Beijing’s official account of what the two sides had agreed upon. While the White House framed the diplomatic dialogue as a serious security issue, listing it alongside traditional nuclear arms control, China’s Ministry of Commerce stressed instead that the talks had taken place, and would continue to take place, strictly under the umbrella of the two countries’ established economic and trade mechanisms. Folded neatly into broader trade talks, a dialogue on artificial intelligence risks and benefits leaves Beijing completely free to parry U.S. requests regarding safety risks—such as placing limits on the open release of models with dangerous capabilities—by demanding that Washington relax its export controls.
To be sure, Xi did propose to "jointly prevent the misuse and abuse of artificial intelligence," conveniently eliding China’s own frequent use of the technology against the United States in various ways, including state-sponsored hacking campaigns and foreign influence operations. He added that artificial intelligence should remain under strict human control and be developed for the public good, a principle Beijing has repeated since 2023 without offering any further detail or concrete follow-through. More pointedly, his insistence that future dialogue should "draw on each other’s strengths, not guard against each other," alongside his explicit request that Chinese companies be "treated fairly," served as clear, implicit rebukes of existing U.S. technological restrictions.
Securing the Advantage
This historical record strongly suggests that Beijing will approach future meetings, beginning with the next round of discussions in November, with proposals explicitly designed to put Washington in a difficult diplomatic bind. Although U.S. Trade Representative Jamieson Greer has made it abundantly clear that export controls are not on the negotiating table, Beijing will undoubtedly raise them anyway. If Washington engages on the topic, Chinese negotiators will claim the existence of a standing diplomatic channel where U.S. restrictions on artificial intelligence technology are open for negotiation. If Washington refuses, Beijing will continue to evade American entreaties on safety while accusing the United States of stifling economic opportunity for the developing world. Beijing deeply hopes that every round of this dialogue will further cement the impression among key U.S. allies—such as Japan and the Netherlands—that Washington is willing to trade away vital technological restrictions over their heads, encouraging those nations to accommodate China’s artificial intelligence ambitions instead.
However, the United States is actually in a substantially stronger position than its current diplomacy often lets on. The world’s leading foundational models are still overwhelmingly American, and, acting alongside its key allies, the United States controls the critical chokepoints in the global semiconductor supply chain. Export controls governing those chokepoints must not be treated as a bargaining chip; rather, they should be actively strengthened. Chinese artificial intelligence companies regularly lament their severe lack of high-end microchips and have frequently resorted to smuggling chips and illicitly distilling American models. Furthermore, U.S. developers conduct increasingly extensive and transparent safety testing, a rigorous practice that Chinese developers largely avoid. Taken together, these distinct strengths give Washington both the necessary means to pressure Beijing and a highly attractive value proposition to offer the rest of the world. The United States should boldly use them, advertising its rigorously tested models to effectively compete for global adoption while simultaneously pressing China hard on safety during bilateral talks.
The first crucial opportunity to apply that strategic pressure is the upcoming bilateral meeting. Washington should establish two firm conditions for this meeting and all future engagements. First, Chinese developers must commit to publishing safety evaluations at least as extensive as those that leading U.S. developers publish today on equivalent benchmarks for the most serious risks, such as cyberattacks and biological weapons, and they must transparently disclose previous and future safety incidents. The United States can readily make the exact same commitments, as American companies already publish such evaluations, increasingly disclose safety incidents, and open their models to independent and government evaluators prior to public release. Without such conditions, the bilateral incident channel agreed upon in September will operate entirely as a one-way street, leaving China completely free to point out U.S. incidents while reporting none of its own.
These binding conditions would fundamentally change the terms of the diplomatic dialogue. Ever since 2023, Washington has raised the issue of safety risks, while Beijing has skillfully steered the conversation toward development. As long as Chinese companies face zero regulatory scrutiny for the risks inherent in their models, Beijing can freely tout them for their openness and innovation without ever having to answer probing questions from the international community regarding their safety. Research by Concordia AI, an organization that tracks China’s domestic artificial intelligence ecosystem, found that only five out of China’s ten leading developers published safety test results for any model over the past year, and none of them did so consistently. If China agrees to these conditions, governments and private companies choosing which artificial intelligence models to deploy will be able to judge Chinese and American developers by the exact same criteria. If China refuses, the onus will rest squarely on Beijing to explain why fair treatment for Chinese companies should not extend to rigorous safety testing. Regardless of the outcome, the U.S. government should ramp up its own technical testing of new Chinese models and actively publish the results, while avoiding unforced political errors—such as past dismissals of artificial intelligence safety concerns as a "hoax"—that hand Beijing an easy retort to legitimate U.S. appeals.
More broadly, effective artificial intelligence diplomacy ultimately requires convincing foreign nations to willingly join the American technological ecosystem. Countries that train their local engineers, build their domestic language models, and run their national data centers on Chinese technology will inevitably come to depend upon it. To effectively counter that dangerous dependence—and prevent it from hardening into long-term political alignment—Washington must present a positive, cohesive vision of artificial intelligence governance that pairs verifiable safety commitments, the kind that Chinese developers consistently avoid, with the actual technological access and financial investment that developing countries desperately desire.
U.S. companies already offer the finest technology in the world, but Washington has yet to put real diplomatic weight behind them. It should draw from its evolving regulatory approach at home to rally international partners around shared safety measures. Under the White House Accord on Super Intelligence, signed in September, six leading U.S. companies committed themselves to rigorous internal controls, external audits, and independent oversight. The G-7 Code of Conduct, which successfully built upon voluntary commitments secured during the Biden administration, could easily incorporate the accord’s stringent measures and serve as a common global standard for trusted international partners. Countries that agree to adhere to these measures could receive priority access to advanced technology and international financing. Without such meaningful incentives, the global artificial intelligence competition may ultimately be determined by price and raw access alone—an arena where China currently holds a distinct edge.
For too long, much of Washington has treated diplomatic talks with Beijing as a neutral space to manage the shared risks of artificial intelligence completely apart from the broader strategic competition. China, conversely, has always treated those same talks as just another vital arena of the larger geopolitical rivalry. To maintain its technological lead and achieve real, verifiable progress on safety, the United States should start doing the exact same thing.
