Pull US AI Research Out of China

Recently, the Biden administration and a host of allies called out China for its massive Microsoft Exchange hack (among others), and threatened strengthened cyber defense measures and continued exposure of the PRC’s malicious cyber activity. But despite a lot of hard talk about securing America’s cyber defenses, action is wanting, and the government has failed to address a glaring boon to the PRC’s cyber capabilities: our own companies’ AI research centers in China.

Housing the AI research labs of America’s cutting-edge tech companies in authoritarian China was never a good idea. But given that the Chinese government uses foreign tech companies to help find and exploit security vulnerabilities, and that it is claiming ever more control over tech companies’ operations and data, it looks more objectionable than ever. AI is an increasingly crucial element of cyber security and hacking, and Xi Jinping’s China has demonstrated time and timeagain that China’s high-tech sector serves the CCP, which sees AI technology in particular as a core tool of its future autocratic rule. 

Nonetheless, according to the Center for Security and Emerging Technology at Georgetown University, 10 percent of the AI research labs of Facebook, Google, IBM, and Microsoft are housed in China. Microsoft’s Beijing-based Research Asia Lab is the company’s largest outside of the U.S. and is credited as being “the single most important institution in the birth and growth of the Chinese AI ecosystem over the past two decades.” The Google AI China Center, which opened just three months before Xi Jinping declared himself dictator for life, includes an elite team of researchers supported by “several hundred China-based engineers.”

That the CCP has an open door to the work and findings of any AI research center in China is well known. Putting aside the Party’s flagrant incursions into foreign firms and the laws that legally oblige tech companies to hand over their data and any recognized security vulnerabilities, the country’s technology sector is built to a considerable degree on its highly organized efforts to acquire technology through underhanded means. 

In 2019, the then-Chairman of the Joint Chiefs of Staff posited that Google’s AI research center in China was “indirectly benefitting the Chinese military,” or perhaps even directly doing so, to the protestations of Google and others. In 2018, Microsoft’s center openly co-authored research with China’s military-run National University of Defense Technology with clear applications to surveillance and censorship.

These research centers are not the first time that America’s tech sector boosted the PRC’s totalitarian governance. Companies like Apple and Microsoft are infamously acquiescent to China’s censorship regime, cutting out access to apps and sites that tell the truth about the Party’s crimes, in addition to providing Chinese citizens’ data to the government. 

Other American companies have gone even further—actively building out the country’s expanding digital surveillance and control architecture for profit. Since Cisco helped lay the foundation for the Great Firewall in the early 1990s and Seagate built the first hard drive catered to surveillance for China’s Hikvision in 2005, American companies laid the foundation for many of the systems powering China’s technological authoritarianism. Their contributions to Xinjiang’s dystopia, such as the Intel chips likely being used to monitor forced labor and concentration camps and Thermo Fisher’s DNA sequencing kits used to surveil Uyghurs, represent the grotesque culmination of this history.

Given that American companies remain the gatekeepers of most of the most valuable insights in advanced AI computing, however, their research efforts within China are disproportionately valuable to the tech-hungry dictatorship, and risky to a world chronically hacked by the PRC. Under the auspices of international scientific collaboration, these R&D outposts ultimately grow the CCP’s capacity to make its own high-tech tools—including for hacking—without having to resort to foreign companies to build out their capabilities.

The danger of China capitalizing on American AI research in its borders also has chilling military import, as our defense leaders know well. Even if the work of these research centers doesn’t have direct application to areas of military concern, the dual-use nature of AI technologies makes secondary military application highly likely, in addition to growing China’s military-pliant AI ecosystem more generally. Success in developing its AI capabilities will further grow China’s leverage and aggression abroad—as if it weren’t already concerning enough.

More comprehensive measures are needed to address China’s hacking campaigns. Legislation that requires tech companies working in China to provide detailed assurances that their personnel, data, and research are not exploitable by the regime to aid in cyber operations or to violate human rights could be a good place to start. So would stipulations barring companies that contract with the Chinese government from also working with the U.S. government on account of security concerns.

But of urgent importance is addressing American companies’ AI research centers in the PRC. International scientific collaboration is a great thing, but not when it is fueling a despotic superpower’s oppression at home and belligerence abroad.

This article was co-authored with my AEI colleague Bill Drexel and originally appeared in Defense One.

Comments (3)
Join The Dispatch to participate in the comments.