On July 17, Xi Jinping stood before the 2026 World Artificial Intelligence Conference in Shanghai and delivered the first keynote address a Chinese head of state has given at the event. The title, Joining Hands to Build a Just and Equitable System for Global AI Governance, stated the argument before the speech began. One day earlier, representatives of 29 countries had signed the agreement establishing the World Artificial Intelligence Cooperation Organization, an independent intergovernmental body headquartered in Shanghai. United Nations Secretary-General Antonio Guterres attended the signing ceremony, although WAICO is not a UN organization. Together, the speech and the agreement moved a campaign three years in the making into a new institutional phase.
China is doing more than promoting a collection of AI policies. It is working to establish the interpretive frame through which governments will decide whose technology is open, whose restrictions are legitimate, whose safety claims are credible, and whose leadership serves the wider world. That is the essence of narrative power. A successful frame does not need to win every argument separately. It organizes how later events are understood, so that each new export control, standards dispute, infrastructure offer, or model release appears to confirm a story already in place.
The mechanism matters because governance language can become strategic infrastructure. Once embedded in institutions, official statements, technical programs, and recurring diplomatic forums, a narrative gains authors, audiences, rituals, and proof points. It begins to reproduce itself. Governments repeat its vocabulary. Officials trained through its programs carry its assumptions into ministries and standards committees. Technology deployments make its claims visible in daily life. Over time, the frame can begin to look less like one country’s argument and more like the natural description of the international system.
Beijing’s core story is now clear. Artificial intelligence is the collective property of humanity. Access to it should not be controlled by a small group of wealthy states or companies. China is prepared to distribute capability, respect the sovereignty and cultural choices of developing nations, and build governance through formal equality among states. The opposing role in that story is reserved for the United States: technologically dominant, increasingly restrictive, and willing to make access contingent on Washington’s security judgments.
The accuracy of that characterization is only part of the issue. Narrative strategy works by establishing the categories through which conduct is interpreted. Under Beijing’s frame, a Chinese capacity-building program becomes evidence of solidarity. An American export control becomes evidence of exclusion. An open-weight Chinese model becomes a public good, while a restricted American frontier model becomes a dependency risk. A state-directed content policy can be described as protection for cultural sovereignty, while American objections can be presented as an attempt to impose one set of values on everyone else.
Xi’s speech organized this story around four themes: openness, security, civilizational diversity, and multilateral governance. The sequence was deliberate. He began with open source, collaboration, and sharing, presenting China as the power distributing the benefits of AI. He then claimed the safety argument, calling for laws, technical monitoring, early warning systems, emergency response mechanisms, protections against malicious use, and systems that remain under human control. Only after establishing China’s commitment to safety did he condemn the ‘overstretching’ of national security and the practice of placing one country’s security above that of others.
That structure is important. Beijing is not arguing that AI risks are imaginary or that governments should abandon controls. It is attempting to define which controls count as responsible governance and which should be understood as technological protectionism. Multilateral rules, technical safeguards, and state oversight are presented as legitimate safety. Unilateral restrictions imposed by the United States are placed in a different category, one associated with exclusion and strategic advantage. The distinction allows China to claim both access and security rather than conceding either side of the debate to Washington.
The conference chair’s statement extended that claim into subjects of direct homeland-security concern. It called for safeguards around frontier models, protection of financial, energy, telecommunications, and transportation infrastructure, defined decision authority and behavioral boundaries for AI agents, mechanisms for behavior tracing and risk alerts, and international action against terrorist, extremist, and criminal misuse. China is therefore positioning itself not only as a provider of technology to the developing world, but also as a source of security doctrine for the systems entering global infrastructure.
Xi’s third theme, civilizational diversity, gives the campaign another important narrative instrument. He argued that AI should preserve the uniqueness of national cultures and respect different histories, social systems, and forms of civilization. That language will appeal to governments that want greater control over how models represent local values, political history, religion, social order, and national identity. It can also provide a principled vocabulary for expanding state authority over model behavior and politically sensitive outputs. Content control is no longer described simply as restriction. It can be framed as cultural protection and sovereign choice.
The fourth theme placed the United Nations at the center of global AI governance and called for an early, consensus-based framework covering development strategies, governance rules, and technical standards. This is where the diplomatic design becomes visible. China’s AI diplomacy has long run through its foreign ministry’s arms-control apparatus and the language of common danger, shared security, and collective management. That language places every state inside the debate as a formally equal participant. A capability contest strongly favors the United States. A legitimacy contest conducted through UN-centered forums gives greater weight to the country that arrives with proposals, funding, institutions, and a moral claim about fairness.
WAICO gives that legitimacy campaign a permanent institutional home. The organization follows the purposes of the UN Charter, but it remains independent of the United Nations. The distinction is important. Guterres’s presence at the signing supplied ceremonial proximity to the UN without converting WAICO into a UN body or constituting formal UN endorsement. For narrative purposes, however, the imagery was valuable: the Chinese foreign minister, representatives of 29 founding states, and the UN secretary-general gathered around the creation of a new organization dedicated to global AI governance.
The signing does not prove that Beijing has achieved global consensus. The founding roster includes many established Chinese partners, but it cannot be dismissed as a narrow authoritarian bloc. Brazil, Indonesia, and South Africa give the organization representation from major democratic and regional powers across the Global South. At the same time, no G7 nation joined, no major European democracy signed, and India, Japan, and South Korea remain outside. WAICO does not yet speak for the world. It does give China a standing platform from which to compete for the authority to do so.
That platform is connected to a wider delivery system. Xi announced 5,000 AI training and seminar opportunities for developing countries over the next five years. He also pledged international AI application cooperation centers with ASEAN, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organization, and BRICS, along with access for 30 countries to China’s AI-powered MAZU meteorological warning system. The organizations he named were not a ceremonial catalogue. Together they form a distribution network spanning much of Southeast Asia, the Arab world, Africa, Latin America, Central Asia, and the expanded BRICS community.
This is where the narrative gains material evidence. Training programs create communities of officials, engineers, and institutional partners whose direct experience reinforces China’s identity as provider. Cooperation centers allow the same language, standards preferences, and policy assumptions to be adapted and repeated across regions. Weather warnings, laboratories, compute access, and practical AI applications give governments something tangible to place beside Beijing’s claim that it is closing the global AI divide. The assistance does not merely accompany the story. It helps make the story appear true.
Open-source AI serves a similar function at the technological level. Xi placed it first among his four observations and described open source, openness, collaboration, and sharing as a historic opportunity. Beijing is presenting open-weight models as proof that China distributes capability while the United States protects its most advanced systems. The appeal is especially strong in markets where cost, local language adaptation, technical sovereignty, and concern about interrupted access matter as much as benchmark performance.
The reality is more complicated than the message. Chinese models vary in openness, hosted services remain subject to Chinese law and political controls, and Beijing is reportedly considering restrictions on overseas access to some of its most advanced future systems. Those contradictions should not be ignored. They also do not erase the strategic value of the open-source frame. Narratives do not require perfect consistency. They require enough visible evidence to organize how audiences interpret competing conduct. A widely available Chinese model can carry more persuasive weight than a diplomatic statement about equitable access.
The campaign’s public record shows how deliberately those proof points have accumulated. Xi announced the Global AI Governance Initiative in 2023. China followed with an AI capacity-building plan, helped move a UN General Assembly resolution on AI capacity building to consensus, launched additional cooperation initiatives, and proposed WAICO in 2025. According to China’s market regulator, Chinese bodies also submitted 505 proposals to international standards organizations in 2025 across fields that included artificial intelligence, biotechnology, and the industrial internet, while China led the formulation and issuance of 285 international standards. The diplomatic narrative and the technical campaign are proceeding together.
Standards are where interpretive authority can become durable technical authority. They define trust assumptions, audit requirements, interoperability rules, safety expectations, authentication practices, and incident-reporting procedures. A country that consistently leads these processes gains influence over the practical meaning of responsible AI. Its preferred concepts move from speeches into reference documents, procurement requirements, and product design. If internationally adopted frameworks increasingly reflect Chinese assumptions about sovereignty, content governance, data control, and state authority, American systems may encounter markets in which their own design choices are treated as the exception rather than the default.
The homeland implications follow from the same process. International standards influence the systems built for global markets, the evidence available to investigators, the provenance mechanisms attached to synthetic media, and the security controls integrated into AI agents and critical infrastructure. Foreign standards do not automatically become American law, but they shape platforms, vendors, and technical ecosystems that cross borders continuously. A governance contest conducted abroad can therefore affect what the United States can trace, authenticate, audit, and defend at home.
The developing world is the campaign’s most important audience. Many governments are selecting their first national AI platforms, building public-sector applications, and deciding which models will mediate education, health, government services, and public information. Chinese-hosted models and turnkey deployments can arrive with political assumptions and content controls embedded in their default behavior, even when open-weight versions can later be modified. The narrative operates twice in these markets. The story of equitable partnership helps determine which technology stack a government selects, and the selected stack then influences the information environment experienced by its population.
A state that supplies another country’s reasoning infrastructure can acquire a form of influence that traditional security and trade relationships do not fully capture. This does not mean every recipient will accept Beijing’s political positions or become dependent on China. Many states will continue to hedge and extract benefits from both sides. It does mean that model behavior, training relationships, standards choices, cloud architecture, and technical support can shape how institutions understand problems long after the original agreement is signed.
The United States is not absent from international AI diplomacy. Washington has built the American AI Exports Program, promoted full-stack technology packages, and assembled 35 countries behind an AI Opportunity Statement linked to the broader Pax Silica effort. Its international story emphasizes trusted technology ecosystems, resilient supply chains, innovation, interoperability, and protection from strategic dependence on adversarial systems. The limited overlap between the American and Chinese groupings suggests that two different political communities are beginning to form around competing definitions of legitimate AI order.
The American strategy has considerable strengths. It is tied to the world’s most capable private-sector ecosystem, leading chips, frontier models, research institutions, and genuine alliances. Its weakness is not the absence of an international program. It is the difficulty of converting a security-centered and commercially structured strategy into a moral narrative that travels as easily as Beijing’s promise of access, equality, and shared development. China offers governments a role in a supposedly common project. American language too often asks them to join a trusted system whose boundaries, technologies, and security judgments are still largely defined in Washington.
Events in June showed how an established frame can exploit that difference. The United States abruptly suspended foreign access to Anthropic’s Mythos-class model because of legitimate proliferation concerns involving advanced cyber capability. The restrictions were later eased, but the interruption gave Beijing a visible example of the dependency risk it had spent years describing. Xi’s warning against overstretching national security addressed the broader architecture of American chip, model, investment, and technology controls. The Mythos episode supplied a fresh proof point that required no explicit reference from the podium.
The same event could be interpreted in two defensible ways. From the American perspective, the suspension was responsible stewardship of dangerous capability. Through Beijing’s frame, it was confirmation that access to American AI can be withdrawn whenever Washington’s interests change. Which interpretation prevails in Jakarta, Nairobi, Brasilia, or Riyadh will not be decided by the technical facts alone. It will depend on which narrative was already available, which institutions repeat it, and which side has built enough credibility for its explanation to feel intuitive.
That is why narrative strategy cannot be separated from technology strategy. The contest is not won by messaging unsupported by performance. Nor is it won by superior performance that others interpret through an adversary’s frame. Beijing is linking language to institutions, institutions to training and infrastructure, and those programs to technologies that can be used as evidence of the story it wants the world to accept. WAICO gives the process continuity. The regional cooperation centers give it reach. Open-source models give it a visible symbol. Standards work gives it the possibility of permanence.
The United States enters this contest with substantial advantages: superior technology, a deeper research base, strong alliances, and a record of making advanced capability available across much of the world. But those advantages will not automatically translate into international influence. Washington must compete in the forums where legitimacy is established, maintain sustained technical leadership in the bodies where standards are written, and give partners confidence that access to American systems will remain reliable even when security restrictions become necessary. It must also explain those restrictions before Beijing explains them first.
China understands that AI governance is not simply a contest over rules. It is a contest over how future restrictions, disputes, security decisions, and technological dependencies will be interpreted. In Shanghai, Beijing moved from constructing that interpretation to giving it an institution capable of sustaining and distributing it. The United States has the capability to lead. The remaining question is whether it will also shape the terms by which that leadership is understood.
