Microsoft’s China ties advanced authoritarian AI, exposing risks of US tech collaboration with adversaries.

In January, when a little-known Chinese AI startup called DeepSeek stunned the tech industry by developing a cutting-edge large language model, one revelation stood out: several of its top engineers had honed their skills at Microsoft’s artificial intelligence lab in China. This was no coincidence and exemplifies a larger trend: Microsoft’s longstanding presence and partnerships in China helped the Chinese Communist Party (CCP) advance its techno-authoritarian ambitions.

Microsoft’s Research Presence in China Fueled a Generation of AI Talent

In pursuit of market access and talent, Microsoft spent decades building research centers and forging collaborations in China. In the process, they trained a generation of Chinese AI experts, transferred knowledge and tools that bolstered China’s military and surveillance state, and complied with the CCP’s censorship demands. As the race for AI supremacy intensifies, it’s time to confront how one of America’s largest tech giants empowered an authoritarian rival’s agenda.

Microsoft first opened its research lab in Beijing in 1998 as part of its larger subsidiary, Microsoft Research Asia. The goal of this lab was to use Chinese researchers to produce groundbreaking research on areas such as computer vision and speech recognition. In the process, Microsoft Research Asia also ended up training thousands of Chinese scientists and engineers, becoming something of a finishing school for China’s tech elite. The founders and top executives of major Chinese tech firms have passed through Microsoft’s labs, including Zhang Yiming of TikTok’s parent company, ByteDance, Tang Xiao’ou of AI giant SenseTime, and leading technologists from Alibaba and Baidu.

Alumni of Microsoft’s Labs Helped Build China’s Surveillance and Military AI Tools

It would be unseemly if Microsoft had merely trained the executives of companies seeking to compete with American companies. But, far more troublingly, many alumni of Microsoft Research Asia have used their expertise to develop technologies that enable oppression, censorship, and genocide. SenseTime, for instance, is a world leader in facial recognition used by Chinese authorities to monitor citizens, especially the oppressed Uyghurs of Xinjiang.

DeepSeek’s Rise Highlights the Limits of Microsoft’s Internal Safeguards

The recent DeepSeek episode is a case in point. DeepSeek’s new open-source LLM achieved viral success and even triggered a $1 trillion sell-off in US tech stocks amid fears of China jumping ahead of the US in AI development. It then emerged that at least four of DeepSeek’s current employees, including a key department chief, previously worked at Microsoft Research Asia’s labs. One worked 10 years at Microsoft’s lab working on large-scale AI model training, and another spent six years there as a “research intern.” Three of the four DeepSeek engineers spent at least five years in Microsoft’s “intern” program developing advanced AI. Geoffrey Cain, an expert on China’s tech sector, noted, “if you’re working for Microsoft for five years at any other job, that’s enough time to climb into a mid-senior or senior position.” These engineers kept the innocuous title of “intern” while acquiring invaluable expertise in AI—expertise they later used to help launch one of China’s most successful AI models.

Microsoft downplays the significance of this talent transfer, insisting that “a handful of former Microsoft interns” were not the secret to DeepSeek’s success. The company noted that it requires employees in China to sign confidentiality and IP transfer agreements to protect sensitive research. They claimed to have set up internal guardrails, such as barring their China-based researchers from working on politically sensitive areas like facial recognition and quantum computing. Yet the DeepSeek case underscores critics’ longstanding fears: even with precautions, a research presence in China inevitably leaks knowledge to Chinese firms. The sheer length of those Microsoft “internships” suggests these individuals weren’t just fetching coffee. They were deeply involved in Microsoft’s R&D and then took that knowledge straight to a Chinese competitor.

US Lawmakers Push Back as Microsoft Begins to Retreat

Amidst the DeepSeek revelations, US lawmakers voiced alarm. President Trump was clear that the sudden rise of DeepSeek’s models “should be a wake-up call” to all American tech companies that “we need to be laser focused on competing.” In a statement accompanying the introduction of a bill seeking to decouple American AI development from China, Sen. Josh Hawley (R-MO) said that “America cannot afford to empower our greatest adversary at the expense of our strength.” Other lawmakers led a bipartisan push to ban the use of DeepSeek’s products on government devices for fear of data siphoning and spying.

For its part, Microsoft appears to have listened to the wake-up call. They have reportedly shut down an AI lab in Shanghai and have begun moving employees from their China arm to Western countries. But this is only the most recent example of troubling entanglement with the CCP.

Beyond talent cultivation, Microsoft has previously collaborated with entities tied to the Chinese state and military. In 2019, it was reported that Microsoft researchers co-authored several AI studies with academics from China’s National University of Defense Technology (NUDT), an elite military-run institution under the Central Military Commission. These papers focused on AI topics, including facial recognition and machine vision. Microsoft portrayed them as typical academic collaborations. But critics noted that the same facial recognition advances could be weaponized for oppression. When this collaboration came to light, then-Sen. Marco Rubio blasted Microsoft’s actions as “deeply disturbing” and “an act that makes them complicit” in China’s repression. As with other dual-use technologies, the line between civilian and military AI applications is blurry, especially in China, where the government openly pursues a strategy of “civil-military fusion” for technological development. By sharing its expertise with NUDT scientists, Microsoft helped China improve tools that facilitate authoritarian control.

In the decades after Nixon’s opening of China, it was postulated that investing in and collaborating with China would result in liberalization. Defenders of Microsoft’s approach might argue that engagement in China had upsides: the company maintained a foothold in a huge market and could claim it was delivering somewhat less biased services than Chinese alternatives. But in practice, that theory has largely broken down. Under President Xi, China has tightened its grip on tech and discouraged the kind of openness that American firms hoped to foster. The flow of benefits has been largely one-way. Microsoft trained experts who built tools for the CCP, published research that China can weaponize, and acquiesced to censorship that strengthens CCP rule, all while Beijing maintained strict limits on US companies and even banned many Microsoft products from government use.

It’s increasingly clear that Microsoft’s presence in China has undercut US strategic advantages and ideals. To be sure, Microsoft is not alone; other US tech companies have faced similar dilemmas. But Microsoft’s deep entanglement makes it a prime example of how American corporate engagement can aid an authoritarian competitor. This calls for a hard reassessment by both industry leaders and policymakers.

It’s Time to Reassess the Costs of US Tech Engagement With China

Microsoft likes to cast itself as a responsible leader in the AI era, investing in ethics and safety. To their credit, Microsoft has taken some steps to wind down operations in China, from Microsoft Research Asia to Microsoft’s first Chinese joint venture, Wicresoft. But living up to that image requires aligning its China practices with its principles. The company is at a crossroads: will it continue business-as-usual in China, supplying advanced know-how and credibility to an authoritarian regime’s tech apparatus? Or will it recognize that some markets come with moral and geopolitical trade-offs that are simply not worth it? Microsoft’s experience should serve as a cautionary tale, prompting a broader recalibration of US tech engagement with China. The message to Microsoft and its peers is simple: stop helping the CCP build the future of authoritarianism.

About the Authors: Luke Hogg and Lars Erik Schönander

Luke Hogg is Director of Technology Policy at the Foundation for American Innovation, where he focuses on the intersection of technological innovation and public policy. Before joining FAI, Luke was Federal Affairs Manager at FreedomWorks, where he concentrated on blockchain, internet governance, and regulatory issues. He holds a BA in Government and Data Science from the College of William and Mary and lives in Washington, DC. You can follow him on Twitter at @LEHogg.

Lars Erik Schönander is a Research Fellow at the Foundation for American Innovation. He was previously a Congressional Innovation Fellow for TechCongress, working for the Senate Committee on Small Business and Entrepreneurship, and a Policy Technologist at the Foundation for American Innovation. He holds a BA from The George Washington University in International Affairs and Economics and is an incoming MBA student at the Tuck School of Business at Dartmouth College. His writing has appeared in the Wall Street Journal, American Affairs, the National Interest, Tablet, and elsewhere.

Image: Shutterstock

The post Microsoft Aided an Adversary’s AI Ambitions appeared first on The National Interest.