미국과의 주요 전략적 경쟁 무대로 부상하는 인공지능을 둘러싸고 중국이 보복 조치를 강화하고 있다.

China is challenging Washington’s attempt to set the rules of the global artificial intelligence race, warning that national-security concerns must not be weaponized to restrict ordinary technological competition. This latest confrontation arrives ahead of expected US-China discussions on AI, with Beijing signaling that any definition of AI security should be jointly agreed upon rather than unilaterally imposed by Washington. A commentary published by Yuyuantantian, a social media account affiliated with Chinese state broadcaster CCTV, emphasized that the distinction between genuine security threats and normal technological competition must be drawn by all sides. This message reflects a deeper shift in the bilateral rivalry: AI is no longer treated simply as a commercial technology. Both nations increasingly view control over advanced chips, AI models, computing infrastructure, and military applications as directly connected to national power.

Beijing’s immediate complaint centers on who gets to define what counts as an AI security threat. The CCTV-linked commentary argued that Washington cannot demand transparency, audits, and safety controls from others while allowing American AI companies to shape those same standards. It also criticized US AI developer Anthropic, accusing the company of problematic data and security practices and insisting that American firms face the same scrutiny that Washington expects from Chinese companies. Anthropic has previously restricted access for Chinese-controlled companies on national-security grounds. Chinese commentary has portrayed moves like this as evidence that the United States is blending safety policy with industrial competition. That disagreement matters significantly, as the US and China are expected to hold dedicated AI talks before President Xi Jinping’s planned visit to Washington on September 24. The negotiations are unlikely to focus solely on preventing dangerous AI systems; they will also address who controls the technologies needed to build them.

Advanced chips sit at the centre of the rivalry. Computing power remains the most important physical resource in advanced AI, as training and running powerful models requires large numbers of high-performance graphics processors and other specialised chips. The United States has spent years restricting China’s access to certain advanced semiconductors and manufacturing equipment. Those restrictions have evolved rather than disappeared. In January 2026, the US Commerce Department changed its policy to allow licence applications for Nvidia H200, AMD MI325X, and similar processors to be reviewed on a case-by-case basis for approved Chinese customers, subject to security conditions. Other advanced-computing exports remain subject to extensive controls. Washington argues that such restrictions are necessary because powerful chips can support military AI, surveillance, cyber operations, and other national-security applications. Beijing sees the same restrictions as an attempt to preserve American technological leadership. That difference sits at the heart of the dispute.

The term “AI arms race” often misrepresents the competition. It does not mean the US and China are simply competing to build an AI-powered weapon first. The contest is much broader, encompassing improvements in intelligence analysis, autonomous systems, battlefield targeting, logistics, cyber operations, drones, and command structures. The same underlying technology can also strengthen civilian industries, scientific research, and economic productivity, making AI a classic dual-use technology. A powerful model developed for civilian applications can potentially be adapted for military or intelligence purposes. This is precisely why semiconductor export controls and AI-model restrictions increasingly sit inside national-security policy rather than ordinary trade policy. The Trump administration itself has repeatedly described competition with China as an AI race, with US officials arguing that maintaining American leadership is necessary for both economic and national-security reasons. China, meanwhile, has made technological self-reliance a strategic priority. Its latest five-year economic roadmap calls for accelerated scientific breakthroughs and wider use of AI across industries, with technology increasingly tied to national security.

The two countries are competing differently. The US currently benefits from leading AI companies, advanced chip designers, cloud platforms, and a large private investment ecosystem. China has built strength through industrial policy, domestic manufacturing, state-directed investment, and rapid adoption of AI across physical industries such as robotics and manufacturing. Brookings researchers have warned that describing the situation as one simple “race” can obscure these differences. The competition is also increasingly creating pressure for technological separation. If the two sides establish different rules for chips, models, data, and cybersecurity, businesses operating globally could eventually face two partially incompatible AI ecosystems. That would have consequences far beyond Washington and Beijing. Countries such as India would have to navigate access to US technology while maintaining commercial links with Chinese hardware, manufacturing, and AI companies.

The next battle is over the rules. The coming US-China discussions could therefore matter even if they produce no major agreement. The two governments first need to agree on basic definitions. What counts as a dangerous frontier AI system? When does an export control protect national security, and when does it become economic protectionism? What information should an AI company disclose to another government? Those questions have no easy answers. But without common rules, every new model, semiconductor, or export restriction risks becoming another point of geopolitical confrontation. For global markets, the US-China AI rivalry could eventually affect which AI products, chips, and digital services are available in different countries and at what price. For India, the bigger issue is maintaining access to advanced technology without becoming overly dependent on either side’s AI ecosystem. The real AI contest is not just about building the smartest model; it is about controlling the chips, standards, infrastructure, and security rules surrounding those models. Washington wants to preserve a technological lead while limiting capabilities it considers dangerous. Beijing increasingly sees those controls as containment. If the two countries cannot agree on where legitimate security ends and commercial competition begins, AI could become another permanently divided technology system, much like telecom equipment and advanced semiconductors already are.