A demonstration of Face++'s facial recognition software at the China Public Security Expo in 2017.Bobby Yip / Reuters

On a recent tour of an enormous, impressive set of artificial-intelligence labs in Beijing, I saw a scene straight out of Silicon Valley: Bright 20-something Chinese researchers with hipster glasses and pink streaked hair sat at row after row of open tables, headphones in, working hard. Projects ranged from innocuous applications like the AI-enabled bicycle share Mobike or the online education portal VIPKid, to ones the U.S. government may find worrying, like Face ++, which is widely believed to be the world’s most powerful facial-recognition software. It can be used to sort photos, but the Chinese security services also use it to ensure they can find any potential “troublemakers” on any street corner in Beijing.

After years of being relatively naive about China’s “all of society” push to catch up and dominate key technology sectors, the U.S. government was jarred awake recently, with the Obama administration sounding alarms about Chinese advances in semiconductors in its waning days, the Pentagon’s Silicon Valley outpost warning about Chinese tech investments in the U.S., and President Donald Trump then taking up a steady anti-China drumbeat.   

Well beyond President Trump and Peter Navarro’s trade war, the U.S. is cracking down. Lawmakers have proposed sweeping expansions to powers of the Committee on Foreign Investment in the United States, the government body that reviews foreign purchases of U.S. assets for potential national-security risks. The legislation would broaden the scope of deals up for review to include all that have foreign investor involvement in so-called “critical technologies” (so far undefined) and non-public technical information. Even before any changes to the law, CFIUS blocked or discouraged approximately 20 Chinese deals in 2017. Members of Congress have asked Education Secretary Betsy DeVos to investigate the Chinese technology company Huawei’s partnerships with more than 50 U.S. universities, and the administration is exploring sweeping visa restrictions for Chinese researchers in science and technology. Investment has responded accordingly: Chinese foreign direct investment fell 35 percent in 2017, while Chinese mergers and acquisitions in the U.S. declined 90 percent in the same time period.

But is the administration solving the right problem? For years, the U.S. was unsuccessful in stopping the Chinese government’s blatant stealing of U.S. intellectual property and its industrial policy that seeks to make the country world-class in key technologies like artificial intelligence and semiconductors—in part for China’s own growing and menacing military to use. The U.S. thereby inadvertently helped China, including its People’s Liberation Army, accelerate its development of these technologies. But in correcting this, the U.S. now risks cutting off too much technology cooperation with the world’s second-largest economy, which would harm U.S. interests more than China’s. The goal should be to protect a few key technologies and come down hard on China’s illegal practices, rather than cutting off technology cooperation generally.

A few examples demonstrate the point:

While Silicon Valley has led the pack in purely digital technologies like social networks, innovation in robotics, drones, and other hardware-heavy technologies requires access to advanced manufacturing. Shenzhen, in China, is the undisputed leader in this. As the journalist Matt Sheehan has argued, Chinese private investors in American start-ups often provide access to this fast-evolving, humming, manufacturing ecosystem—and as one U.S. entrepreneur put it, a week working on manufactured product design in Shenzhen is worth four in the U.S., because one can tinker with physical designs so much more quickly.  

Google and Microsoft are two of the world’s leading AI research entities. Recently, they have both been criticized on Fox News and elsewhere for opening research facilities in China, because, as the argument goes, some of the innovations made by their Chinese researchers will “leak” to the Chinese government. That is indeed possible, and those companies should put strong controls on employees and invest in cybersecurity to prevent this. But severely curtailing this kind of cooperation, which would be implied as the logical end point of these critiques, makes no sense. Even if U.S. and Chinese researchers can’t collaborate, Google has a large AI team in Europe, which the U.S. government cannot prevent from working with its Chinese colleagues.

For their part, Chinese technology companies like Baidu and NIO happily locate research centers in Silicon Valley, a practice from which the American economy benefits in both insight and jobs.

This kind of cooperation is only going to grow more important as technology gets more complex. As the analyst Paul Triolo notes, the upcoming buildout of 5G telecom networks will require supply-chain contributions from a variety of countries—including China—and restricting work with China in this area will only serve to undercut American competitiveness. “Going it alone” in spaces like 5G would also mean that the world would adopt someone else’s (likely China or the EU) governance standards, which would be bad both for U.S. consumers and for the security America is belatedly trying to protect.

Finally, long-term innovation happens because U.S. companies participate in markets all over the world. A senior U.S. semiconductor-company executive told me recently that U.S. companies’ sales to and investments in China give them access to Chinese efforts on cloud computing, AI, and autonomy, and thus make U.S. products much better at responding to a changing world market. Balkanization will likely harden the Chinese desire to be self-reliant, and so lead to inefficiency, higher costs, and less innovation among U.S. firms. He added that to prepare for more U.S. CFIUS and export control scrutiny, U.S. firms will have to spend additional money on compliance red tape, which could be better used for research and development.

The Chinese AI lab I visited perfectly demonstrates the problem—and benefit—of interconnectedness. Its founder is a Taiwanese national, educated in the U.S., with deep ties both to the U.S. and China. Many U.S. investors participate in his funds. Most of the technology they incubate is innocuous—making it easier for all of us to share bikes in cities or to educate our children. Some of it helps the Chinese government spy on its own citizens. Should U.S. investors be prohibited from investing in his funds so no American capital flows to Face++? Should he be prohibited from investing in U.S. companies, even though the money he is investing comes in large part from U.S. investors?

Many U.S. analysts are writing about a new “space race” in AI. But this Cold War analogy isn’t the right one. In the case of the current technology competition between the U.S. and China, there aren’t two nicely defined “blocs” of countries pitted against each other on behalf of national “teams.” Instead, a web of interconnected universities, researchers, supply chains, and capital flows take the best microchips from America and Korea, embed them in the best hardware from China, and use AI researchers of all nationalities who often work far from their country of origin and publish their research for all to see on open-source platforms.

Chinese intellectual-property poaching is a genuine problem, and U.S. government officials are right to work valiantly to stop it, and to slow the technical advances of the Chinese military. Yet without being clear about the problem they are solving, they risk overcorrecting and stifling positive cooperation on benign technologies, while imposing more burdens on U.S. companies.

If the concern is illegal technology transfer to the Chinese government and military, then U.S. regulation can be both more narrow, and more broad. More narrow in that the U.S. can carefully scrutinize Chinese investment that seeks access to U.S. technology that helps the Chinese military (rather than discouraging all Chinese investment), and carefully vet Chinese students to make sure they are not intelligence assets, rather than effectively discouraging all of China’s best minds from studying here. In fact, the U.S. could encourage them to stay by issuing H-1B visas to the most promising grads, and thus taking advantage of this Chinese resource. Broader, in that countries from Europe to Japan and Korea are also concerned about Chinese technology mercantilism and theft, and America can lead a united front against it.

Building a wall to keep America’s technology in and others’ out is not the right analogy—rather, America should snip the few bad strands in the otherwise positive web of international tech innovation.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.