China’s theft of American AI tech is becoming more brazen

China’s artificial intelligence progress remains heavily dependent on theft and smuggling of American technology, despite the hype surrounding Beijing’s AI capabilities. The Chinese Communist Party is determined to maintain tight control while pursuing what Xi Jinping calls an ‘epoch-defining technology’ – and appears confident that President Trump has limited options to counter increasingly brazen activities.

This escalating tech theft comes as Trump and Xi prepare for a Beijing summit this week, with the Chinese leader betting that America’s preoccupation with the Iran war leaves little bandwidth for a tech confrontation that could reshape global power dynamics.

Last month, the White House accused Beijing of ‘industrial-scale’ theft of know-how from American AI labs. US prosecutors claim to have busted an international smuggling ring that funneled advanced chips worth billions of dollars to China in defiance of sanctions. Meanwhile, Beijing blocked a $2 billion takeover by Meta of Chinese AI startup Manus and prevented the company’s founders from leaving the country.

The theft accusations center on a process called ‘distillation’, where China illicitly trains smaller AI models on the output of larger, expensively developed US models. A leaked internal memo from Michael Kratsios, director of the White House Office of Science and Technology Policy, stated:

“The US government has information indicating that foreign entities, principally based in China, are engaged in deliberate, industrial-scale campaigns to distill US frontier AI systems.”

The distillation process involves creating thousands of fake accounts for targeted AI chatbots or tools, with accounts working together to extract information. US AI company Anthropic detected 24,000 fraudulent accounts that generated more than 16 million exchanges with its Claude chatbot. The company accused leading Chinese labs of running the campaign to acquire powerful capabilities ‘in a fraction of the time, and at a fraction of the cost’.

Anthropic warned that ‘distilled’ apps would carry none of the original safeguards against using AI for developing bioweapons or cyberattacks, ‘creating significant national security risks’. This represents a new front in technology espionage that bypasses traditional hacking methods.

Beijing has also established an extensive smuggling network to circumvent US restrictions on top-end Nvidia chips used for training AI models. Federal prosecutors describe how servers containing ‘billions of dollars’ of restricted chips were shipped to front companies in Southeast Asia before being repackaged and diverted to Hong Kong and mainland China. One surveillance video showed a defendant using a hair dryer to swap sticky labels and serial number tags – a bizarrely low-tech method for high-stakes smuggling.

The indictments represent just the tip of a chip-smuggling iceberg in what Xi characterizes as ‘a race to the top’. This struggle for AI supremacy goes beyond economic competition to a battle that will define the future balance of global power.

Xi’s determination extends to keeping AI firmly under Chinese Communist Party control, as Meta CEO Mark Zuckerberg discovered. His takeover of Manus seemed like a done deal, with employees already moved into Meta’s Singapore office. Manus creates AI agents that act as autonomous personal assistants for tasks ranging from product launches to stock market analyses.

Chinese authorities didn’t specify which laws the deal violated, but the move sends a clear warning to Chinese AI startups against taking technology outside China. ‘Beijing effectively drew a bright red line that Chinese AI talent and technology are not for sale to American companies, full stop,’ Han Shen Lin of The Asia Group told Reuters.

By torpedoing the deal, the CCP killed the practice of ‘Singapore washing’ – where Chinese tech companies like Shein and TikTok shift headquarters to Singapore to appear less Chinese. All Chinese companies are legally required to assist Beijing’s intelligence agencies, making such geographic moves largely meaningless.

A recent US-China Economic and Security Review Commission hearing warned that China is harvesting US data to build ‘AI-enabled intelligence and targeting architecture for economic competition, political coercion and wartime advantage’. Despite strong bipartisan Congressional support for aggressive export controls, Trump is sounding more dovish ahead of the summit.

Xi likely calculates that Trump has limited options and little appetite for trade hostilities that could disrupt the global economy further. Last year, Trump scaled back tariffs after Beijing weaponized rare earths, threatening to restrict access to critical minerals over which China holds a near monopoly.

Even as evidence of AI-related theft grows, Trump has sent mixed messages. He eased controls on some Nvidia chips and suggested openness to Chinese automakers building vehicles in the US despite security concerns. His harsh words for allies contrast sharply with his friendlier approach to Xi, recently writing on Truth Social:

“They have agreed not to send weapons to Iran. President Xi will give me a big, fat, hug when I get there… We are working together smartly, and very well! Doesn’t that beat fighting.”

By last year’s end, around a third of AI models downloaded worldwide were Chinese. Xi is riding a wave of ‘Chinamaxxing’ – online infatuation with Chinese technology driven partly by Western influencers courted by the CCP. Themes of China’s innovation prowess and greater societal AI acceptance are widely promoted by credulous Western analysts.

The reality is more complex, given Beijing’s continued reliance on large-scale theft. Chinese company DeepSeek startled the world with AI models performing almost as well as Western ones at a fraction of the cost. But its latest release was met with indifference – not only more expensive to build but reportedly subject to far more CCP meddling.

Humanoid robots performing kung fu and competing in marathons provide entertainment value, but experts question real-world applications. They represent an extravagantly wasteful state-led program that mirrors Chinese innovation more broadly.

Security and control remain Xi’s overwhelming priorities. While unleashing spies to harvest know-how and chips, Chinese-developed algorithms must align with party doctrine. The CCP frets about safety, cybersecurity, and job impacts. The recent freezing of 200 robotaxis that clogged Wuhan’s streets provided one wake-up call.

With youth unemployment around 17% officially, young people struggle to find jobs. The CCP worries about ‘tangping’ (‘lying flat’) – young people dropping out of high-pressure job markets for simpler lives. Last month, China’s Ministry of State Security declared tangping a foreign conspiracy to undermine Chinese youth. While the AI race heats up, the CCP’s basic control instincts remain chillingly familiar.