_________

Published on Aug. 18, 2026

Should AI be treated like a scarce resource such as oil, and priced accordingly? Or should it be considered a public good, like water, that belongs to all of humankind, including the poor and all people in the "Global South"?
Staying with this metaphor for a moment, Chinese President Xi Jinping has clearly positioned China as a proponent of the water approach to AI. Speaking at the recent World Artificial Intelligence Conference (WAIC) in Shanghai, Xi said that all countries "should take a people-centered approach and develop AI for the positive and for good".

Open-source AI models are the answer, both the Chinese president and his country's official AI strategy are arguing. "As a new engine of world economic growth and an accelerator for the shift of growth drivers, AI is moving from the digital world into the physical world. We should seize this rare, historic opportunity to encourage open-source, openness, collaboration and sharing", Xi Jinping said in his keynote speech at WAIC on July 17, 2026, the official Chinese news agency Xinhua reported.

The Chinese leader promised to help developing countries gain better access to AI, to make, in his words, "this frontier technology better benefit humanity". At the same time he repeated his earlier proposal for "global governance" to deal with the risks of AI. "We should practice true multilateralism and recognize the important role of the United Nations", Xi said.

In the run-up to the conference, the People's Daily, the Communist Party's mouthpiece, had already made it clear that Beijing wants to be seen as an alternative model to the U.S., where the government is trying to gain dominance in AI via export controls against China and where a leading AI company, Anthropic, bans Chinese developers from using its products.

The U.S. was not explicitly named in the People's Daily. It was clear anyway that it was the target of all comparisons. "Two fundamentally different approaches to AI development currently coexist. Some media have described them as an “oil mindset” and a “water-flow mindset.” The former regards data and computing power as scarce and exclusive resources, encouraging barriers, blockades and zero-sum competition. The latter views artificial intelligence as an inclusive public good, encouraging connectivity, sharing and mutually beneficial cooperation", the newspaper wrote.

It is obvious that China is now treating open-source AI and widely diffused open-weight models from the likes of DeepSeek, Kimi or Alibaba not only as a commercial good, but as a vehicle of soft power. As models trained in China start answering user queries around the world, Chinese values might one day become as universally distributed as American values are today thanks to Hollywood and its dream machine, or so Beijing believes.

Of course, soft power is not only about technology or ubiquity, but also about trust. Already, some users of Chinese models complain that answers on sensitive political topics seem to reflect Beijing's censorship or self-censorship on the part of the Chinese model houses.

This argument, however, cuts both ways. If every nation is likely to infuse its cultural values and, arguably, its many biases into the frontier models it trains, getting your own models out there becomes critical in the long run.

Andrew Ng, the AI pioneer who has been both cofounder of Google Brain and chief scientist at the Chinese technology company Baidu, and who is today one of the most influential thought leaders on AI, calls it a "brilliant move" that Chinese companies have released many open-weight models, Noema magazine recently reported.

"Do we want stories told about the importance of liberty and democratic values, or do we want stories told that reflect other nations’ values? When someone asks a question of AI, such as what happened in Tiananmen Square in 1989, whose model they are using will cause the model to give an answer that reflects that nation’s values. So getting your stuff out there is a tremendous source of soft power influence,” Ng recently told an audience in Los Angeles.

Interestingly, both the Chinese and U.S. governments now publicly advocate open-source AI as part of their national strategies. The U.S. plan, Winning the Race: America's AI Action Plan, encourages small and medium-sized businesses at home to adopt open models.

“The document cites the importance of creating open models “to ensure America has leading open models founded on American values” and because open models’ potential ... lends them “geostrategic value”, writes the think tank RAND.

The nonpartisan think tank, which is largely financed by the U.S. government, has come to the conclusion that Washington should do even more to further the existence of homegrown open-source or open-weight models, in order not to fall behind the Chinese competition. It cites a U.S. Center for AI Standards and Innovation finding that DeepSeek downloads on model-sharing platforms rose nearly 1,000 percent after the R1 release of January 2025. Among its recommendations for Washington are a stronger U.S. open-model ecosystem, recalibrated export controls and incentives for permissive licensing.

In other words, experts in the U.S. recommend that their own government should do exactly what the Chinese are already doing. In reality, however, it has become common knowledge that Chinese AI developers are currently far ahead in releasing powerful open-weight models, which are cheaper to use and enable a growing number of companies around the world to run AI tasks on their local servers and save on inference costs, including many Silicon Valley start-ups.

Chinese developers such as DeepSeek, Alibaba, Zhipu and Moonshot have largely embraced the open-source framework. Their models are free to download, and anyone may inspect and adapt them.

The giveaway is "turbo-charging their global adoption and boosting usage," Bloomberg recently reported.

Recently, two new Chinese AI models, Moonshot's Kimi K3 and Alibaba’s Qwen3.8-Max-Preview, have made international headlines because of their powerful capabilities, trailing only the best American models by a very narrow margin on certain benchmarks.

"Because they are open-source, anyone moderately skilled in the new tech can create bespoke programmes, unlike the restrictive use imposed by AI providers in the US", wrote the South China Morning Post in Hong Kong. "China is catching up with the United States in the AI race more quickly than previously estimated. Its AI models charge a fraction of the fees levied by their US rivals", the paper commented, reporting that the releases have been called "a second DeepSeek moment".

For AI developers in many countries and also for international companies embedding AI in more and more of their business processes, the censorship of certain political topics may be less important than the lower token cost of Chinese models, plus the ability to adapt open-source or open-weight models on their local servers, so that their sensitive company data doesn't have to be shared with anyone.

On the importance of open-source AI, Chinese President Xi Jinping recently found an unexpected ally in Meta Platforms Inc.'s Mark Zuckerberg. In a 6,500-word essay published on August 10, 2026, he argued that the industry's future depends on broader access to AI models.

“The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic,” Zuckerberg wrote. Observers interpreted it as challenging the closed vision of AI championed by companies like OpenAI and Anthropic.

Zuckerberg's company invested heavily in its open-source Llama family of models, which, according to Bloomberg, “ultimately underwhelmed Silicon Valley.” Over the past year, Meta pivoted and prioritized a new generation of models that it could keep proprietary.

While he now suddenly championed open models again and announced the release of Meta's new open-weight system called Muse Glimmer, he nonetheless argued for continued U.S. technology boycotts intended to limit China's progress in AI. It seems openness should absolutely have its limits, as far as Zuckerberg is concerned. Boycotts of advanced hardware holding down Chinese AI competitors are very welcome.

Part of China's embrace of the open-source route in AI stems from the need to play defense against such U.S. attempts to contain its growing economic and military power. Just like chipmakers in the People's Republic are embracing RISC-V open-source technology to circumvent restrictions on advanced U.S. chips and their corresponding specifications, open-source AI models offer an alternative pathway that leads away from American dominance.

Chinese experts generally acknowledge that large AI models from the U.S. are currently still superior, although that gap is shrinking fast. They see their country's biggest chance in establishing an early lead in physical AI, where increasingly capable large models meet a huge and rapidly modernizing industrial supply chain.

Open-source or at least open-weight models facilitate the work of Chinese engineers who experiment with a vast number of practical deployments of vertical models, and they also enable a faster diffusion of new automation solutions across whole industries.

That is why Xi Jinping stressed that AI is currently moving from the digital world into the physical world and called it a "rare, historic opportunity". Chinese officials and AI developers are less interested in frontier technologies like artificial general intelligence (AGI) than their American counterparts, and more interested in practical AI use cases in manufacturing, biotech or renewable energy.

The oil versus water comparison used by the People's Daily is rooted in ancient Chinese philosophy. It should not be mistaken for one of the cheap propaganda swipes at American profiteering for which the paper is otherwise known. Rather, it explains the very different thinking behind the Chinese and the American approaches to AI.

All educated Chinese know a sentence from the Daodejing, written by Laozi. "Shang shan ruo shui, shui shan li wan wu er bu zheng" can be translated roughly as "The highest good is like water. Water benefits all things and does not compete."

Artificial intelligence, therefore, is understood in China as something that should not be treated as a commercial commodity alone, and not even just like any other public good. It is something that lifts all boats and ultimately resists attempts to harness it for narrow competitive endeavors, be it by individual companies or by individual nations.

In his WAIC speech in Shanghai, the Chinese president quoted another proverb. "We often say in China, "A single string cannot make music, and a single tree does not make a forest." AI development should not be a solo performance by a single country, but a symphony of international cooperation", he said.

China has just made a determined attempt to position itself as a responsible, inclusive AI superpower. Critics will undoubtedly call it propaganda, but it is also a powerful vision. It will resonate in many countries where people wonder if they might one day also be on the receiving end of U.S. technology boycotts.