Publications

The Chinese Vision: Xi Jinping's Speech

In July of this year, Chinese President Xi Jinping delivered an address for the first time at the World Artificial Intelligence Conference (WAIC), lending the event particular political and strategic significance. Xi opened by referencing the 1956 Dartmouth Conference, where the term "artificial intelligence" was coined. By choosing an American milestone as his starting point, Xi acknowledged the United States’ historical contribution while simultaneously presenting the development of artificial intelligence as a global process in which China now seeks to play a leading role. Xi compared artificial intelligence to the revolutions brought about by the steam engine, electricity, and the internet, illustrating that AI is a technology that will reshape the economy, society, and the international system, and therefore demands new governance mechanisms as well.

This concept integrates into China's broader foreign policy which, since 2021, has introduced the Global Development Initiative (GDI), the Global Security Initiative (GSI), and the Global Civilization Initiative (GCI). All of these reflect China's ambition to influence the shaping of international institutions, norms, and governance rules.

Xi outlined four principles for global governance of the AI field: technological openness and the promotion of open-weight models, which allow users to download models, run them independently, and adapt them to their needs; safety and risk management, coupled with implicit criticism of the expansion of national security considerations in the United States to justify export restrictions; respect for cultural and value diversity among countries; and multilateralism under UN leadership, alongside assistance to countries in the Global South in capacity building.

Xi also presented practical steps aimed at demonstrating Chinese commitment: training 5,000 artificial intelligence experts from developing countries over the next five years; establishing cooperation centers with BRICS, ASEAN, the Shanghai Cooperation Organisation (SCO), the Arab League, the African Union, and the Community of Latin American and Caribbean States (CELAC); and making an AI-based meteorological early-warning system accessible to 30 countries. In doing so, China presented itself not only as a developer of technology, but also as a country seeking to provide capabilities, knowledge, and infrastructure to its partners.

WAICO: The Institutional Framework of the Chinese Strategy

The ideas presented by Xi took practical form in the announcement of the establishment of the World AI Cooperation Organization (WAICO). The organization follows China’s established pattern of creating international institutions under its leadership and extends it into the field of artificial intelligence. Headquartered in Shanghai, the organization is based on the Global AI Governance Initiative, introduced by China in 2023, and initially comprises 29 founding member states, most of them from the Global South. Its objective is to create a permanent framework for cooperation in artificial intelligence through workforce training, joint research, the development of standards, efforts to narrow digital divides, and expanded access to AI technologies. WAICO reflects a familiar pattern in Chinese foreign policy: the establishment of international institutions and frameworks under Chinese leadership, as seen in the Asian Infrastructure Investment Bank (AIIB) in the field of finance, and the Belt and Road Initiative (BRI) in infrastructure. In this sense, WAICO is not an end in itself, but a mechanism designed to expand the adoption of Chinese technology and deepen its influence.

It is still too early to determine whether the organization will indeed become an influential international institution. A study by Tsinghua University examining the global AI governance map, argues that WAICO's potential distinctiveness lies in its combination of openness in principle to countries wishing to join, without setting conditions based on regime type or political characteristics, and an emphasis on development and narrowing gaps in AI capabilities. Alongside this, the researchers emphasize that the organization is still in an initial stage: its operating mechanisms, budget, and membership structure have yet to be established. Its success will therefore be tested in the coming years by its ability to transform from a political declaration into an active international framework that attracts countries, resources, and influence, and shapes the environment in which it operates.

Two Competing Strategies for Shaping the Global AI Order

The United States and China are currently pursuing two different strategies for shaping the global AI order, each seeking to build an international bloc around itself through distinct mechanisms: the United States relies on control over core technologies and value chains, while China relies on accessibility, training, infrastructure, and multilateral institutions. The US strategy is based on maintaining its technological advantage through control over the industry's critical value chains: advanced chips and manufacturing equipment, computing infrastructure, and leading models. The export restriction policy that began during the Biden administration, and was subsequently expanded, was designed to slow China's progress by limiting its access to the most advanced technologies. Simultaneously, initiatives such as Pax Silica seek to strengthen cooperation between the United States and its allies in the fields of semiconductors, computing infrastructure, supply chains, and research in order to preserve the Western bloc's technological edge.

China, on the other hand, promotes a different strategy built on four complementary components: developing competitive models at low costs; disseminating open-weight models; expanding collaboration, training, and infrastructure with countries of the Global South; and establishing international institutions, chief among them WAICO, that will provide a permanent framework for cooperation.

WAICO and Pax Silica illustrate two distinct paths to forming international coalitions in the field of artificial intelligence. China's success will be measured largely by the number of countries, organizations, and developers that adopt the models, standards, and frameworks it promotes.

The Chinese Models: The Technological Foundation of the Chinese Strategy

For WAICO to become an influential framework, China must complement its institutional dimension with genuine technological capabilities. In recent years, GLM, Qwen, DeepSeek R1, and ERNIE have demonstrated that China’s AI industry is capable of consistently producing advanced models that compete at the global level, despite US export restrictions on advanced chips. Kimi K3, developed by the Chinese Moonshot AI and launched on the eve of the AI conference in July, stood out for its competitive performance and its distribution as an open-weight model. The model illustrates the advantage China seeks to offer: performance that, in some tasks, approaches that of leading US models, combined with open weights and competitive usage costs. This combination may be particularly attractive for companies, universities, and countries seeking to reduce costs and maintain greater control over how the model is operated.

However, alongside the technological achievements, the model also sparked controversy over its origins: it has been alleged that Kimi K3 was developed in part through distillation from Anthropic’s model. In this method, a new model learns from the outputs of a more advanced model. The method itself is widely used in the industry, but using a closed model without authorization raises intellectual property questions. Moonshot AI denied the allegations, and skeptics pointed out that the short interval between the launch of the Anthropic model and the launch of Kimi K3—only about two weeks—makes it difficult to claim that it indeed served as a central basis for training. The controversy itself indicates that the competition between China and the United States is no longer limited to model performance, but also extends to questions of ownership, origin, and the rules that will govern AI development.

Alongside their advantages, the use of Chinese models raises questions regarding the transparency of training data and development processes, data security, and the risks associated with their integration into sensitive systems. Although open weights allow models to be run independently and can reduce some dependence on the provider, they do not eliminate the need to assess each model and its operating environment individually.

From Beijing's perspective, the importance of models such as DeepSeek and Kimi extends beyond their technological achievements. They allow China to offer a practical alternative to the American system. China is not required to lead in every performance metric to establish global influence. Models that deliver strong performance alongside low costs, open weights, and broad accessibility may be sufficient to achieve widespread adoption. As countries, universities, and companies expand their use of these models, they may build applications on top of them, train personnel, and develop local capabilities. Widespread adoption of the models may create an additional advantage: as more users and developers operate within an environment built around Chinese models, the costs of switching to alternatives increase, strengthening the incentive to continue using solutions from the same ecosystem. Over the long term, this trend could make it difficult for Western alternatives to gain a foothold in markets where Chinese models have already achieved broad distribution. In this way, Chinese models will become the foundation of a broader technological system, granting China influence even without absolute superiority in performance.

Implications and Recommendations for Israel

For Israel, the competition between China and the United States in artificial intelligence is expected to affect its foreign policy, security, supply chains, and innovation ecosystem as the two powers continue to build competing frameworks for cooperation, standards, and infrastructure. Access to advanced AI models is increasingly being treated as a strategic resource, subject to national security considerations, export policies, and access restrictions.

Israel's technological edge relies on its strengths in innovation, cooperation with the United States, and access to chips, computing infrastructure, and research. Its integration into initiatives such as Pax Silica is therefore consistent with its security and technological interests and its strategic alignment with the Western bloc.

At this stage, Israel has no interest in joining WAICO. The organization is still in its infancy, its operational mechanisms have yet to take shape, and joining it could create friction with the United States and complicate Israel's integration into Western technological frameworks. WAICO may offer access to infrastructure, models, research, and partnerships that are attractive to Israel, but these benefits must be weighed against the strategic implications of joining a framework led by China in a field at the heart of great-power competition. This case differs from Israel's decision to join the Asian Infrastructure Investment Bank (AIIB) in 2015, when competition between the great-powers was less intense and infrastructure financing was not as central to technological and security competition as artificial intelligence is today. An additional challenge concerns Chinese models in the Israeli market. Their advantages in terms of price, performance, and accessibility could encourage widespread use, gradually increasing reliance on the development environment and applications built around them.

Three policy recommendations follow:

First, Israel must deepen cooperation with the United States and other partners within the Pax Silica framework in chips, computing infrastructure, research, and standardization, in order to maintain access to technologies and resources essential for the development of advanced artificial intelligence.

Second, Israel should regularly monitor WAICO’s activities, the standards it develops, the countries that join it, and the partnerships China advances through the organization. This will enable Israel to assess WAICO’s international influence and adjust its policy accordingly.

Third, Israel should formulate a national policy for the use of foreign AI models. Such a policy should include risk assessments prior to deploying models in government systems, critical infrastructure, and defense organizations, as well as tracking the extent of their adoption across the economy and the ecosystem developing around them, in order to identify security risks and technological lock-in.

Alongside these measures, Israel should promote AI literacy among users in both the public and private sectors in order to maximize the benefits of new technologies while reducing exposure to their risks.