To Win the AI Race, America Must Remember Why It’s Worth Winning

Every powerful technology eventually poses the same question: will it expand freedom or suppress it? AI is forcing the United States and our democratic allies to confront a hard reality: we are in a close contest with China, the world’s foremost authoritarian power, and whoever shapes AI will shape the 21st century.

Open-source AI is not so open

At this year’s World AI Conference in Shanghai, Chinese leader Xi Jinping called open-source AI, which is freely available and increasingly dominated by Chinese companies, a “rare historic opportunity” to shape the technology’s development on China’s terms. A recent Reuters investigation found Chinese researchers are extracting data from leading American AI models to build systems for surveillance, cyberwarfare, drones and other military applications.

American AI is fueling China’s military machine.

White House science adviser Michael Kratsios has described Chinese activities as industrial-scale extraction campaigns. Treasury Secretary Scott Bessent was more blunt: “Open source is not open season on American IP.”

The Chinese Communist Party’s policy of Military-Civil Fusion erases the line between civilian technology and military use, including in AI. Whether obtained legally or not, NVIDIA chips, U.S. intellectual property, and distilled American AI models all end up serving the People’s Liberation Army.

Palantir CEO Alex Karp, in The Technological Republic, quotes philosopher Michael Sandel: “Fundamentalists rush in where liberals fear to tread.” That’s what happened with rare earths, the minerals powering modern weapons, jets, and electric vehicles. When America’s last major rare‑earth mine closed in 2002, China seized the opening and built the processing capacity no one else would. Today, its control of roughly 80 percent of the supply chain is a serious vulnerability.

From rare earths to AI

Chinese open-weight AI models are following the same pattern. Alibaba’s Qwen alone has logged more than three billion downloads in six months. Yet many technologists and investors hyping open-weight AI ignore the difference between legal American models and stolen Chinese ones.

Technology isn’t values neutral. It reflects those who build, deploy and control it. Chinese AI reflects the values of the Chinese Communist Party. CrowdStrike recently found that prompting DeepSeek-R1 on topics Beijing considers sensitive, like Tiananmen, Uyghurs, or Taiwan, raises the odds of severe security vulnerabilities in its code by as much as half. Instead of building the digital infrastructure of free societies on an undemocratic foundation, we should aggressively back American open models that carry our own democratic values, so countries deploying AI at scale have a real alternative that preserves their agency. That’s a matter of principle. It’s also a matter of strategy.

Today, the U.S. is ahead of China, “but we are not comfortably ahead,” says Keith Krach, Chairman of the Krach Institute for Tech Diplomacy at Purdue. Epoch AI, a nonpartisan research group, estimates that our lead is only seven months. Once AI is advanced enough to accelerate its own research, today’s modest advantage could either become decisive or vanish.

74 technology races

The warning signs extend beyond AI. The Australian Strategic Policy Institute’s tracker of 74 critical technologies found China leading in 69 of them in high-impact research. That matters because AI isn’t one thing. It’s a stack: hardware, data systems, AI models, and applications. The country whose tech stack the world adopts will have enormous influence over the future.

There are two ways to win the AI race. One is to build the most capable systems. The other is to make your technology so widely adopted that it becomes the default global infrastructure.

China could lose the frontier race and still win if its “good-enough” systems become entrenched across enough of the world.

We have seen this before. Huawei and ZTE showed how cheap Chinese technology could embed itself in other nations’ 5G infrastructure, creating lasting dependencies it took a U.S.-led Clean Network campaign to unwind. The same pattern is playing out in AI. Developers across Africa are already choosing Chinese open-source models, The New York Times reported, and once those models become the foundation for e-government systems, education tools, and critical infrastructure, the dependencies will be far harder to reverse.

So what should the free world do?

First, stay ahead at the frontier. It makes no sense to supply our top adversary with the most advanced capabilities it needs to surpass us. We must protect America’s lead in advanced chips and semiconductor manufacturing while closing loopholes that allow computing power to leak to China.

Second, outbuild China. We need far more compute, energy, and semiconductor capacity so that America’s advantage compounds faster than China can catch up.

Third, win the adoption race. Simply telling Africa, Latin America, the Middle East and Southeast Asia to reject Chinese open-weight AI will not work if the only American alternative is closed and expensive. To that end, we need to incentivize high-performing American open-weight models that developers abroad can actually build on. According to former Nigerian Minister Oby Ezekwesili, Africa doesn’t just want to be told what to reject. It wants a seat at the table in building the technology that shapes its own future.

The White House’s America’s AI Action Plan recognizes this, calling for exporting the full American AI stack to allies and partners. Through the National Champions Initiative, the Trump Administration has invited tech companies from allied countries into its AI export packages.

Now, it’s time to execute.

No democracy can produce the entire AI ecosystem alone. Taiwan’s TSMC makes the world’s most advanced chips. The Netherlands’ ASML makes the machines that produce them. Japan supplies critical semiconductor equipment. Israel contributes cybersecurity. And American companies such as Nvidia and AMD lead in chip design.

Autarky breeds redundancy. The real goal is building resilience by aligning strategic capabilities among trusted allies: joint research, coordinated export controls, harmonized procurement, shared compute, and workforce training.

Indispensability is the real prize. Whoever provides the technology the world runs on will help shape the values embedded in that technology.

Beijing spent decades building “socialism with Chinese characteristics”: censorship, surveillance, and unquestioned obedience to the state. It is now extending those characteristics into AI.

That’s why the engineers, entrepreneurs, and innovators building AI across America and the free world have a responsibility beyond making it more capable. Because they are helping determine what kind of world that capability creates, they should build AI with the characteristics of a democratic society.

The future will run on AI. Let’s make sure it runs on freedom.

This article was originally published by RealClearDefense and made available via RealClearWire.