China is rewriting the ‘Soft Power’ playbook for the age of AI

During the Cold War, “soft power” meant American movies and music that won hearts and minds abroad. Today, in the digital age, China is pushing a very different kind of soft power: its open, low-cost artificial intelligence technology.

In a speech last week, Chinese leader Xi Jinping identified so-called open-source AI as key to China’s ambitions to shape the technology’s global development. Mr Xi hailed open-sourcing, in which much software can be freely shared and modified, as a “rare and historic opportunity” to spread the benefits of the technology around the world.

Mr Xi has made it clear that China will push for an open path as an alternative to relying on US tech giants, whose AI software is typically closed and more expensive.

“The development of artificial intelligence should not be a solo performance by one country,” Mr. Xi said, “but a symphony of global cooperation.”

The Chinese are calling for a new AI order amid a growing battle for global dominance between top US companies and fast-moving Chinese rivals.

US companies such as OpenAI, Anthropic and Google are keeping their technology close, arguing that guarding the underlying code ensures greater security and control over powerful AI systems. Chinese companies such as DeepSeek, Moonshot AI, Z.ai and Alibaba have been pushing open source AI. They argue that these systems cost less, will accelerate global adoption and are no less secure.

Promoting open-source AI “serves China’s twin goals of strengthening its image as an inclusive AI leader and reducing global dependence on US AI,” said Kyle Chan, a fellow at the Brookings Institution who is an expert on China’s technology policy.

As Chinese companies rapidly close the gap in AI performance, the country’s rivalry with the United States has become more hostile. US companies have accused some Chinese competitors of improperly extracting data from their systems to help develop their own technology.

Last month, Anthropic sent a letter to two US senators accusing Chinese tech giant Alibaba of a “reckless” and “illegal” attempt to copy its technology using 24,000 fraudulent accounts. Alibaba has denied any wrongdoing.

The Trump administration takes such allegations seriously. On Wednesday, Treasury Secretary Scott Bessent threatened to impose sanctions on Chinese AI companies.

“We support open-source AI and the innovation it unlocks,” Mr. Bessent wrote on X. “But open source is not open season on American IP,” referring to intellectual property.

American tech executives and investors also worry that China’s open-source technology could change the economics of AI production. Their models are more efficient, they consume less computing power. This innovation is partly out of necessity, as US trade restrictions have limited access to Nvidia’s most advanced AI chips.

If China’s AI alternatives are nearly as good and cheaper — estimates range from 50 to 90 percent for some tasks — they could threaten the business plans and sky-high valuations of America’s AI leaders. And Chinese technology is increasingly being adopted by engineers and companies in the United States in an effort to cut costs.

China’s strategy is “to gain market share around the world, even if it doesn’t immediately translate into profits,” said Wendy Chang, a senior analyst at the Mercator Institute for China Studies think tank.

Open access advocates point to Linux, an open source operating system, as reason for optimism. This software runs on most computers in Internet data centers and benefits from many people examining the code and making fixes.

Still, computer experts agree that artificial intelligence is a more subtle case. It includes more than just code, especially the data used to train the model. And AI models make judgments and can reason.

There are also security concerns. Some security experts insist that the open-source approach releasing both the software and the numerical settings for AI training models, known as weights, greatly increases the risk that the technology will be used for cyberattacks or even to help make biological weapons.

And the Chinese government’s strong influence over its companies raises potential data protection and security concerns.

But the closed models of Silicon Valley companies have their own security flaws. This week, OpenAI revealed that a model it was testing had gone bad and attacked another company’s computer systems. The target company, Hugging Face, said it deployed a Chinese open-source model from Z.ai to analyze the data and counter the attack.

Now China has outlined a plan to use its open source advantage as a diplomatic tool.

The concept of open-source AI software as a soft power tool has been popularized in technology circles. It suggests influence through persuasion and desire—something people want—rather than the coercive “hard power” of military and economic power.

The Open Path offers a low-cost foray into AI for software engineers and for nations looking to capture the next wave of technology.

“Those who benefit from free AI software certainly recognize that,” said Lawrence Lessig, a professor at Harvard Law School and an expert on technology policy. “From a political point of view, this is a huge gain for China.”

China, an authoritarian power, often uses the language of openness in reaching out overseas, especially in the developing world. And in recent years, it has used AI in these efforts, creating joint AI development and training centers in dozens of countries. This year, China established the World Organization for Artificial Intelligence Cooperation to shape AI management principles and promote the adoption and training of AI in China-friendly countries.

In his speech, Mr. Xi promised more investment. This included 5,000 AI training programs and seminars for students and workers in developing countries over the next five years, as well as new or expanded AI initiatives in Africa, Asia and Latin America.

On the same day as Mr. Xi’s speech, the People’s Daily, the official newspaper of the Communist Party, published an article about China’s AI investments abroad, with a litany of training and joint development programs.

His congratulatory conclusion: “China is helping countries around the world, especially those in the Global South, to strengthen their capabilities and gain wide international recognition and support.

Xinyun Wu contributed reporting from Taipei, Taiwan, and Alan Rappeport from Washington.