“When it comes to emerging technology, you cannot be both in China’s camp and our camp.”
Commerce Secretary Gina Raimondo described the global competition for supremacy in artificial intelligence (AI) this way last April. She’s right, and the Biden administration has taken important steps to restrict China’s access to advanced AI capabilities, from banning the sale of advanced chips last year to formulating restrictions on the export of certain AI models.
But these steps miss another important and overlooked problem: American Big Tech companies also threaten U.S. national security by providing technology to entities linked to the Chinese Communist Party (CCP). The U.S. government must take concrete steps to rein in Microsoft, Amazon and other tech firms that conduct business with these firms in ways that harm our national security.
The computing power required to train artificial intelligence models is inherently a “dual use” technology. It has many harmless commercial purposes, but China has seized on it for cyberwarfare and other military applications.
How does China access such resources? While entities can no longer buy advanced AI chips directly from American companies like NVIDIA, China can bypass these sanctions by purchasing them remotely via the cloud services sold by Microsoft, Google and Amazon.
Raimondo recognized this gaping hole in our sanctions. “We use export controls on chips,” she said, “but what good is that if they go around that to use our cloud to train their models?”
American Big Tech companies have long shown that they won’t voluntarily prioritize U.S. national security over profits. They must be forced or pressured to do the right thing.
Microsoft is one company that has recklessly engaged CPC-connected companies. Late last year, The Verge reported that ByteDance, TikTok’s Chinese parent company, had been licensed by Microsoft to use OpenAI’s technology to train their own AI model, violating the terms of service. Microsoft Azure cloud services that enable AI training are available to other Chinese clients who successfully apply for or are otherwise granted access, including 3,900 startups in Hong Kong as well as startups in mainland China via Microsoft’s Accelerator program.
Amazon is also entangled, as Amazon Web Services provides machine learning technology and training through its Chinese affiliate to entities in China that contract with its military, the People’s Liberation Army (PLA).
These companies also enable tech transfer to the CCP with their physical presence in China. Microsoft runs an AI research center in Beijing and employs 10,000 people in China — at least 3,000 focus on AI development. This is a risk to American security, as China’s intelligence laws require any Chinese national to partake in state intelligence operations, even if their employers don’t allow them to spy for the government.
Clearly, something is wrong. Microsoft is reportedly asking hundreds of employees who work on AI and cloud computing in China to relocate. But the company says it remains “committed to China,” and a handful of voluntary relocations will hardly protect Americans’ privacy and security while Microsoft maintains its massive Chinese footprint.
U.S. tech companies have long sought to downplay the significance of their business dealings in China. They claim to have guardrails in place to prevent intellectual property theft or nefarious use of technology. But voluntary corporate guardrails are toothless, given how Chinese intelligence laws and civil-military fusion laws are structured. Private American companies cannot stop their partners and license buyers in China from handing their dual-use technology to the PLA.
Microsoft’s lobbying to the Commerce Department shows its naivety about China. It suggested that the department impose controls that target specific “entities of concern” as opposed to the whole country. However, in passing the TikTok bill, the Biden administration and Congress, informed by enormous evidence, rejected claims that ByteDance was supposedly independent from the CCP. Washington should also dismiss any hollow and unprovable idea that American companies can export only to “the good guys” in China while eschewing bad actors.
What’s the solution? The Commerce Department should expand its efforts to block American cloud computing services from helping the Chinese military apparatus. The department should thoroughly investigate and expose security risks from Big Tech’s cloud sales and AI research in China and impose legal requirements of transparency over these activities.
These are realistic goals; the U.S. government has enormous leverage over these companies through lucrative, taxpayer-funded federal contracts enjoyed by Microsoft, Google and Amazon. The government can attach strings to those dollars: you cannot sell cloud computing services to any hostile authoritarian state, including China, because our enemies will turn those services against us.
Finally, American tech companies acquire small firms at a breakneck pace, opening another avenue for reform. Regulators at the Department of Justice and Federal Trade Commission can use their merger review process to open the books into these companies and investigate how newly acquired firms will integrate into the American parent companies’ business in China.
The message: If you are an American company, you fly the American flag, not China’s. Cloud computing is a significant and dangerous source of revenue for American companies in China. It must be decoupled.
Geoffrey Cain is policy director at the Tech Integrity Project, nonresident senior fellow at the GeoTech Center of the Atlantic Council and author of “The Perfect Police State: An Undercover Odyssey Into China’s Terrifying Surveillance Dystopia of the Future.”