Despite efforts by the United States and allied governments, China’s appetite for foreign technology and use of extraordinary means to obtain it has not diminished. At the same time, China’s capacity for indigenous research, evidenced in the recent DeepSeek algorithmic AI breakthrough, has made collaboration between the U.S. and China less of a one-way street.
These conflicting dynamics present U.S. policymakers with a dilemma: how do we protect scientific innovation—arguably our greatest asset—from exploitation by China without closing useful channels of cooperation? In other words, how do we have our cake and eat it too?
There can be no question that China’s state-sponsored transfer practices include illegal activities. Its methods for illegal tech acquisition run the gamut from cyber theft, cloak-and-dagger espionage, and insider operations to patent and copyright infringement, reverse engineering, and violation of NDAs. Losses run in the billions.
These illegal practices, moreover, divert attention from even more damaging acquisitions. China also sponsors venues to co-opt foreign owners of technology without host countries’ knowledge—a practice referred to by insiders as “non-traditional collection.” Beyond the ubiquitous “talent” programs, this strategy includes targeting databases, parallel labs, “short-term” visits, overseas support guilds, dedicated transfer offices, fronts, and a network of commercialization centers.
Worrisome as these extralegal transfers are, they pale in comparison to what China gains from legal practices, such as China-based multinational research, tech spotting and talent scouting, startup competitions, and company buyouts—not to mention direct and indirect tech purchases, and technology-for-trade arrangements.
So why allow these practices? Why invite exploitation, in China’s terms, “by various means?” Is the risk worth the gain? Our answer is a qualified “yes.” To understand why, consider the field of artificial intelligence (AI).
China has already achieved near-peer status in most areas of AI, outguns the United States in applications, and is on track to achieve parity or dominance by 2030. Efforts to “slow down” China by chip embargoes push its scientists toward creative solutions, as the DeepSeek models amply demonstrate. Given this state of affairs, there are real gains to be made through collaboration or, to be totally cynical, real opportunities to reverse or even subvert the transfer process.
Chinese AI talent leads the world in numbers of published papers, conference presentations, AI graduates, and patents. More than a third of the AI researchers in the United States hail from China. Their contributions to U.S. research far outweigh the inevitable leakage.
Notably, China’s notion of how general AI should be pursued differs from ours—and we can learn from it. The U.S. is immersed in a monoculture of AI development centered on large generative models, whose disadvantages are legion and future far from guaranteed. China is less wedded to a single paradigm and invests heavily in alternative approaches. Engagement with China would give us better insight into how a competitor is proceeding.
Anthony Kwan/Getty Images
And of course, with engagement comes leverage. AI safety is a universal problem that demands some form of global governance in which China participates. There is a genuine cadre of ranking scientists in China concerned about AI’s impact on human welfare and willing to work with their Western counterparts to address the problem.
We remember the Cold War as a time of clarity, and many hope the U.S. will imitate it in its approach to China today. But today’s challenges are more nuanced. China is not like the former Soviet Union in terms of population, economic strength, global reach, and market impact. Managing technology loss to China by sealing all doors would be the epitome of “cutting off one’s nose to spite one’s face.”
That said, the present circumstances—whereby China “borrows” freely from the West while pursuing its independent programs—cry out for mitigation. The following recommendations can protect U.S. technology while preserving the incentives that attract foreign talent.
First, recognize that the venues and organizations that enable China’s transfer operations are known or knowable entities. Given China’s need to attract participants, there is a limit to what can be hidden from law enforcement and motivated analysts. These data should be aggregated and shared with academic and industrial compliance offices, who themselves would benefit from federal funding.
Second, set clear legal guidelines on what is and is not permissible, and communicate these strictures broadly within the United States and among non-U.S. citizens contemplating research or study here. Scientists caught in illicit transfer operations often do not know what they are participating in, let alone their degree of legality. Clarity and forewarning will reduce anxiety among our guests and help eliminate concern over arbitrary enforcement.
Third, encourage skilled foreign researchers to remain in the U.S. by fast-tracking citizenship, conditioned on their abiding by common-sense policies on respecting proprietary technology. “Loyalty” can be enhanced by salaries and conditions that at least match what China offers its researchers—after all, people go where they are valued.
Fourth, require members of China’s technical support guilds, sponsors of and participants in China’s “talent” recruitment programs, U.S.-based technology spotting firms, and others associated with China’s United Front and technology acquisition network to declare their status under the Foreign Agents Registration Act. The United States could even incentivize elements of this network to work on America’s behalf.
Fifth, compel recipients of U.S. government contracts to report to their compliance offices all professional contacts with or travel to China and other designated nations, ideally in advance. These data should be made available to analysts tracking China’s legal, illegal, and extralegal transfer operations. Meanwhile, eliminate stovepipes that inhibit information sharing between tracking agencies.
Finally, understand that these temporizing measures are no substitute for national policies aimed at growing America’s science and technology base to avoid a zero-sum struggle.
William Hannas is the lead analyst at Georgetown University’s Center for Security and Emerging Technology (CSET), a former member of Georgetown’s Asian Studies faculty, and previously served with the U.S. Navy and the CIA’s Senior Intelligence Service. Huey-Meei Chang is a senior China S&T specialist at Georgetown University’s Center for Security and Emerging Technology (CSET) and formerly a U.S. Department of Defense and CIA contract instructor for Chinese document exploitation.
The views expressed in this article are the writers’ own.
