Microsoft has been working with a Chinese military-run university on artificial intelligence research that could be used for surveillance and censorship.
Three papers, published between March and November last year, were co-written by academics at Microsoft Research Asia in Beijing and researchers with affiliations to China’s National University of Defense Technology, which is controlled by China’s top military body, the Central Military Commission.
Samm Sacks, a senior fellow at the think-tank New America and a China tech policy expert, said the papers raised “red flags because of the nature of the technology, the author affiliations, combined with what we know about how this technology is being deployed in China right now”.
“The [Chinese] government is using these technologies to build surveillance systems and to detain minorities [in Xinjiang],” she added.
The US government is currently debating whether research collaborations, particularly in sensitive areas such as artificial intelligence and augmented reality, should be subject to stricter export controls.
Adam Segal, director of cyber space policy at the think-tank Council on Foreign Relations, said: “US-China academic partnerships are increasingly under the microscope as the FBI focuses on the threat of espionage from students and scientists, and the defence department [focuses] on the possibility that frontier technologies might eventually make their way to the PLA [People’s Liberation Army].”
Last week, the Massachusetts Institute of Technology cut ties with telecoms group Huawei and launched an “elevated risk” review process for its Chinese collaborations.
Microsoft said its “researchers, who are often academics, conduct fundamental research with leading scholars and experts from around the world to advance our understanding of technology”.
It added: “In each case, the research is guided by our principles, fully complies with US and local laws, and . . . is published to ensure transparency so that everyone can benefit from our work.”
One of the papers co-authored by Microsoft and researchers affiliated with the NUDT described a new AI method to recreate detailed environmental maps by analysing human faces, which experts say could have clear applications for surveillance and censorship.
The paper acknowledges that the system provides a better understanding of the surrounding environment “not viewed by the camera”, which can have a “variety of vision applications”.
Pedro Domingos, a professor of AI at the University of Washington, said: “Let’s suppose I’m an intelligence agency and I have pictures of people of interest; I can use the system to tell something about the place they’re in that they didn’t realise they were giving away.”
Microsoft’s research arm in Beijing has collaborated on at least two other papers with NUDT researchers, including in the area of machine reading — a way for computers to understand online text.
“Machine reading comprehension may not seem directly concerning but it could be used for censorship, which is an interest of the Chinese government,” said Elsa Kania, a Chinese military technology expert and a fellow at the think-tank Center for a New American Security. “They see natural language processing as a way to enable censorship at scale.”
The papers underscore Microsoft Research’s long-running links to Chinese military-funded academia, including its operation of several “tech clubs” for students at Chinese universities known to have military links including NUDT, Beihang University and Harbin University of Technology.
The US commerce department is currently seeking public comment to help identify emerging technologies that are essential to US national security because they may enable intelligence collection, weapons of mass destruction, or terrorist applications, or could provide the US with a qualitative military or intelligence advantage, according to law firm White & Case.
“[These papers are] at the centre of a perfect storm with . . . issues top of mind as the US is redrawing its relationship with China,” said Ms Sacks.
“Firstly, there is a massive overhaul of the export regime, and emerging and foundational technology, particularly AI, is a considered category under that. Because this is open academic research, the question is to what extent is the Chinese government exploiting the openness . . . and extensive history of joint partnerships,” she added.
Additional reporting by Christian Shepherd in London and Richard Waters in San Francisco