The U.S. government gave at least $30 million in federal grants for research led by a scientist who is now at the forefront of China's race to develop the most advanced artificial intelligence—which he compared to the atomic bomb due to its military importance, a Newsweek investigation has revealed.
Pentagon funding for Song-Chun Zhu, the former director of a pioneering AI center at the University of California Los Angeles (UCLA), continued even as he set up a parallel institute near Wuhan, took a position at a Beijing university whose primary goal is to support Chinese military research, and joined a Chinese Communist Party "talent plan" whose members are tasked with transferring knowledge and technology to China.
Newsweek's revelations underline how the United States, with its open academic environment, has not only been a source for China of advanced technology with military applications but has also actively collaborated with and funded scientists from its main rival. Only as tensions with China have grown over everything from global flashpoints to trade to technology has the research started coming under growing scrutiny.
Responding to Newsweek's questions over funding for Zhu, the Department of Defense said there were also advantages to international collaboration: not least being able to recruit top minds from around the globe, including China, to the United States.
The National Science Foundation, a federal agency that was among those which awarded millions of dollars in grants to Zhu, began to use new analytics tools to fully determine potential conflicts of interest in 2022.
"The foreign collaborations and affiliations of Song-Chun Zhu were identified and reported to the intelligence community and law enforcement," Rebecca Keiser, chief of Research Security Strategy and Policy at the foundation, told Newsweek. "The NSF became aware of these national security and research security risks near the end range of this scientist's funding," Keiser added.
Newsweek has seen no accusations that Zhu broke any U.S. law. He did not respond to several emailed requests for comment and nor did his research institute. China's embassy in Washington did not respond to requests for comment.
To build a picture of the funding and of Zhu's activities, Newsweek examined federal grants databases, scientific papers, reports from Chinese and U.S. universities and companies, and Chinese local government announcements.
Among key donors were the specialist military research agency DARPA, the Navy and the Army. The Department of Defense for years has highlighted the essential role that artificial intelligence is expected to play in the battlefields of the future.
Most of the federal grants awarded to Zhu, a professor of statistics and computing, were in the decade before 2020, the year he returned to China after 18 years at UCLA's Center for Vision, Cognition, Learning and Autonomy. But two grants included the year 2021: one was for $699,938 to develop "high-level robot autonomy" that was "important for DoD tasks, such as autonomous robots, search and rescue missions," according to the Department of Defense's grants website. Another, for $520,811, aimed to build "cognitive robot platforms" for "intelligence and surveillance systems via ground and aerial sensors." Zhu was named as principal investigator on both grants awarded by the Office of Naval Research.
"China has built a vast system to extract technology and know-how from US federally funded research," said Jeffrey Stoff, director of the Washington, D.C.-based Center for Research Security & Integrity, a non-profit set up to mitigate the risk to research from U.S. adversaries.
"Compounding this problem is double-dipping, where recipients of federal research dollars like Zhu also take Chinese government funding for the same research and divert efforts for China's benefit," Stoff said.
China's leaders openly say they aim for the People's Liberation Army to surpass the U.S. and allied militaries in technology and capabilities. AI is a core part of that, with China aiming to outdo the U.S. in multiple spheres of power including economic and geopolitical by the mid-century, as Chinese leader Xi Jinping pursues what he calls "changes unseen in a century"—and strengthens ties with other U.S. rivals such as Russia and Iran.
As U.S.-China relations worsened under the administration of former President Donald Trump, Chinese researchers in the U.S. began to come under greater scrutiny. In 2020, Trump ordered a ban on graduate students from China with ties to the military. President Joe Biden has also shown concern over China's acquisitions of U.S. technology, imposing restrictions on the transfer of some sensitive items such as AI capabilities.
Underlining the importance of the issue, Biden issued an executive order on October 30 with the intention "to ensure that America leads the way," as well as establishing greater safety and security standards.
Beijing has consistently denied stealing technology secrets from the United States and other countries.
Underlining Western concerns, the heads of the "Five Eyes" intelligence-sharing network of the U.S., Canada, Britain, Australia and New Zealand made an unprecedented joint appearance in mid-October to warn that massive technology extraction by China was putting the U.S. and the West at a military, economic and security disadvantage, threatening its leadership in science and innovation.
"What if Nazi Germany or the Soviet Union had won the nuclear race instead of the United States? We might have lived in a very different world," Condoleezza Rice, the director of the Hoover Institution at Stanford University and a former secretary of state, said at the spy chiefs' event.
'Atomic Bomb' of Information Technology
Zhu was upfront about his goals after returning to China in 2020 to found the Beijing Institute for General Artificial Intelligence (BIGAI). In a proposal submitted in March to a top political body he belongs to, the Chinese People's Political Consultative Conference, Zhu called the most advanced AI "the strategic commanding heights of international scientific and technological competition in the next 10 to 20 years."
"Its influence is equivalent to the 'atomic bomb' in the information technology field. If our country can take the lead in realizing a truly universal intelligence, it will become the 'winner' of the international technological competition," Zhu said.
In a statement on its website, BIGAI said that "military, industry, social governance, and cognitive confrontation" were the core of an "emerging strategic field where countries are competing for deployment."
While scientists in the West warn that artificial general intelligence could run amok and destroy human civilization—with some calling for a total moratorium on development—there is less public debate in China, where Xi Jinping says the country should focus on realizing it, and is investing to do so.
Zhu is also chief scientist at the Wuhan Institute for Artificial Intelligence, founded last December in the city which is at the center of a separate controversy over U.S. research funding for a virology lab there, following the emergence of the coronavirus in Wuhan.
A Defense Department official, who declined to be identified, confirmed many of Zhu's grants to Newsweek but said, "He was involved in awards with multiple investigators and used only a small portion of the awards cited," adding that "these are awards to the institutions and reflect multiple researchers rather than an individual."
The Pentagon highlighted what it said were advantages to international collaboration, such as the ability to attract and retain the best talent from around the world. It said about 90 percent of Chinese students who complete a science or technology graduate degree in the U.S. stay on. The U.S. and the Department of Defense also benefited from open academic research that spurred inspiration and groundbreaking ideas, the defense official said.
The National Science Foundation, which awarded 13 grants worth at least $5 million to Zhu as principal or co-principal investigator—for work on computer vision and cognition which are seen by Chinese researchers as core to potential human-like or super human-like intelligence—said it had grown concerned over Zhu's links after publications, collaborations and overseas positions showed conflicts of interest.
"We were looking for researchers who were members of talent plans, who had formal professional affiliations that were not disclosed, funding that had not been disclosed, and undisclosed foreign patents," Keiser said. The NSF did not specify when exactly it alerted U.S. intelligence and law enforcement. The FBI declined to comment.
UCLA confirmed that Zhu was principal investigator on grants awarded to the university totaling approximately $22 million over the last 15 years.
In a statement, UCLA spokesperson Ricardo Vazquez said that Zhu retired from the university in October 2022, two years after returning to China, and that he currently had no external funding from any source through the university. However, he retains the title of professor emeritus, and remains doctoral supervisor of students—many from China— still at the center, according to its website and to their online biographies.
"Foreign engagement is of fundamental importance to UCLA," Vazquez said in an email, adding: "UCLA understands the importance of protecting the scientific enterprise and cooperates fully with law enforcement and our federal funding partners when they conduct investigations."
Significant Future Impacts
Experts say that technology loss through research can be difficult to measure as it may have significant military or economic impacts only in the future.
"Zhu Songchun and his activities are a great example of the complexity of this problem," Stoff said, using Zhu's original Chinese name. "He is an expert in AI disciplines deemed of critical import to the U.S., which is why he received so much Department of Defense and other federal research funding."
"However, I question what the U.S. government gained from its extensive investment in Zhu since he established a research institute and companies in China whose key personnel were trained by him as PhD students and postdocs," Stoff said.
Zhu is not the only high-profile, U.S.-trained AI scientist to return to China in recent years, according to the Center for Security and Emerging Technology, a Washington, D.C., think tank. Others include Harry Shum, Microsoft's former executive vice-president; Pu Muming, who taught at top U.S. universities including Berkeley; and Yao Chi-chih, who taught at MIT and received the Turing award in 2000.
Nor was Zhu the only researcher with undisclosed Chinese links to have attracted the attention of security and law enforcement agencies. Concerned about security, in 2022, a federal agency examined a sample of six universities and found three to five research security breaches in each, Newsweek understands.
Increasingly, the authorities are asking for their money back.
In early October, Stanford University agreed to pay back $1.9 million after the Department of Justice said it failed to disclose foreign support for 12 faculty members.
Among them, Newsweek research showed, was prominent U.S. chemist Richard Zare. He was principal investigator for research that attracted grants for $3.8 million from the Army, the Air Force and the NSF, even as he took money from the National Natural Science Foundation of China and was a member of the Fudan University High-Level Talent Initiative, which like other talent plans also includes some foreigners.
Stanford responded on behalf of Zare to a request for comment.
"We are pleased to have resolved this matter and remain firmly committed to supporting our researchers in meeting federal compliance responsibilities as they pursue their important work," spokeswoman Dee Mostofi said, adding, "Stanford takes seriously the threat of foreign governments seeking to undermine U.S. national security."
In another high-profile case, another U.S. chemist, Harvard University's Charles Lieber, was sentenced in April to time served and fined for lying to federal authorities about his affiliation with China's Thousand Talents Plan, having an undeclared, paid position at the Wuhan University of Technology, and failing to report income as required by law.
Kevin Gamache, chief research security officer at the Texas A&M University, who is prominent among a group of academic office-holders trying to address the problem, told Newsweek that international collaboration was foundational for research, but that the benefits needed to balanced against the risks.
"We need to understand who we are collaborating with, who is funding those collaborators, and whether there is a state-sponsored nexus for that collaboration." Gamache said. "Within the U.S. research community there is still a lot of denial that a problem exists."
A Parallel Life
Zhu had a long association with the United States.
He obtained his doctorate at Harvard University in 1996 after undergraduate studies in computer science in China, going on to work at Brown, Ohio State and Stanford universities. In 2001 he won the Navy's young investigator award—today worth up to $750,000—as well as a NSF career award, worth $340,000 that year. In 2003, he won the prestigious Marr prize for computer vision.
But he was also accumulating positions in China.
In 2004 Zhu founded the Lotus Hill Institute of Computer Vision and Information Science in his hometown of Ezhou near Wuhan, together with Shum who is now described by Microsoft as an emeritus researcher.
Summers spent working in China and year-round collaboration with scientists there was the norm, according to Chinese-language state media and institute reports.
In 2010, Zhu became the first person from Ezhou to be appointed a member of the Thousand Talents Plan of the Communist Party, the party-building website of the local government reported proudly. The plan aimed to bring home "strategic scientists and leading scientific and technological talents who can break through key technologies, develop high-tech industries, and promote emerging disciplines," it said.
The Thousand Talents Plan is the best-known of hundreds of programs in China designed to transfer technology and knowledge from overseas.
Zhu was also a doctoral advisor and a researcher at the Beijing Institute of Technology, one of China's civilian universities known as the "Seven Sons of National Defense," which have the stated mission of supporting China's military research and defense industries.
Many of his students from UCLA joined Zhu at BIGAI in Beijing.
Among them was Liu Hangxin, team leader at the Robotics Lab, where work on similar projects to UCLA continues for example into the "cognitive architecture" underpinning intelligent human behavior and how to transfer it to robots with the aim of creating autonomy via perception, reasoning and motion planning, according to online biographies at github.io and BIGAI's website.
And research by Zhu and his former students and colleagues that was co-funded by federal grants continues to be published today.
One paper from 2022 aims "to disentangle the appearance and geometric information" of image and video data to identify "color, illumination, identity, category." Its lead author, Xing Xianglei, is at Harbin Engineering University, another of the "Seven Sons" military-affiliated universities. Zhu and the current professor of statistics at Zhu's former UCLA center, Ying Nian Wu, and other students and colleagues, are identified as co-authors. The paper credits DARPA and the NSF for funding, as well as the National Science Foundation of China and the Natural Science Fund of Heilongjiang Province of China.
Didi Kirsten Tatlow can be reached at d.kirstentatlow@newsweek.com. Follow her on X, formerly Twitter, at @dktatlow
fairness meter
To Rate This Article
