By Michael Zhuang

The White House has issued a sharply worded memo accusing China-linked entities of systematically stealing U.S. artificial intelligence (AI) technology on an “industrial scale” and vowing to hold them accountable.

The warning, ahead of President Donald Trump’s planned visit to China, underscores the sharpening pace of U.S.–China competition, as officials and analysts say the race for AI dominance is rapidly becoming a matter of national security—and increasingly a form of strategic confrontation.

A Rare, Direct Accusation

In the April 23 memo, issued by White House science and technology policy chief Michael Kratsios, U.S. officials said they had evidence that foreign actors—primarily based in China—were carrying out coordinated efforts to extract sensitive capabilities from leading American AI systems.

According to the memo, these entities used tens of thousands of proxy accounts to evade detection and deployed jailbreaking techniques to bypass safeguards, allowing them to systematically extract proprietary information and replicate functionalities from U.S. models.

The Trump administration pledged a range of countermeasures to combat what it described as the large-scale, illicit exploitation of American innovation.

Entering a New Phase

Analysts say the language used by the White House marks a turning point.

Sun Kuo-hsiang, a professor of international affairs and business at Nanhua University in Taiwan, told The Epoch Times that labeling the activity as “industrial-scale” suggests Washington no longer views such actions as isolated hacking incidents or corporate misconduct.

Instead, he said, the White House is framing the issue as an organized, repeatable system—one that effectively compresses the time and cost required for China-based firms to develop advanced technologies by stealing from the United States.

A similar view was echoed by Hsieh Pei-hsueh, a research fellow at Taiwan’s Institute for National Defense and Security Research, who told The Epoch Times the characterization signals a “qualitative shift” in how the United States assesses China’s technology acquisition strategies.

“This indicates that U.S.–China competition has entered a new phase—one that will be more public and more confrontational,” he said.

Hsieh compared the importance of AI to that of nuclear weapons or energy resources.

“Future economic systems, military capabilities, and decision-making processes will all depend on AI,” he said. “Computing power and model weights are becoming strategic assets—like oil, weapons, or even nuclear capabilities.”

China’s Denial

China has denied the allegations.

On April 24, Chinese Foreign Ministry spokesperson Guo Jiakun said at a routine press conference that the accusations were “completely unfounded” and that Beijing “firmly opposes” them.

However, U.S. officials and lawmakers point to a growing body of cases they say illustrate the problem.

Earlier this year, AI startup DeepSeek drew scrutiny over allegations it used “distillation” techniques—training smaller models on the outputs of larger ones—to build competitive systems at lower cost.

In February, American AI firm Anthropic accused several Chinese firms, including DeepSeek, Moonshot AI, and MiniMax, of attempting to extract capabilities from its models.

Broader Threat and Countermeasures

Concerns over intellectual property theft by communist China extend beyond AI.

At an April 22 hearing titled “Stealth Stealing: China’s Ongoing Theft of U.S. Innovation,” the Senate Judiciary Committee heard testimony warning that China has systematically targeted American advances in fields ranging from biotechnology and energy to telecommunications and weapons systems.

Senate Judiciary Committee Chairman Chuck Grassley (R-Iowa) said in his submissions that China systematically steals U.S. intellectual property through tactics such as corporate infiltration, litigation abuse, and technology transfer schemes, causing massive economic losses and posing a growing threat to American innovation and national security.

In response, U.S. policymakers are pursuing a mix of legislative, legal, and technological measures.

Rep. John Moolenaar (R-Mich.), chair of the House Select Committee on China, has co-sponsored legislation with Rep. Michael Baumgartner (R-Wash.) aimed at preventing foreign adversaries from extracting key features from proprietary U.S. AI systems.

Meanwhile, criminal prosecutions are beginning to emerge. In January, former Google engineer Linwei Ding, a Chinese national, was indicted in federal court in San Francisco for stealing AI-related trade secrets and selling them to China-based companies.

Sun said broader controls may be on the horizon.

These could include expanded export restrictions, not only on semiconductors but also on AI-related assets such as APIs, model weights, and cloud computing access. Closer coordination between the federal government and major tech firms—including OpenAI, Anthropic, Google, and Microsoft—is also expected to deepen, according to Sun.

Looking ahead, Hsieh warns the global AI ecosystem could become more restricted.

He said that companies may limit the capabilities of open-source models to reduce the risk of extraction, while governments could impose stricter identity verification requirements—such as real-name registration or biometric checks—to prevent the mass creation of fake accounts.

At a broader level, the debate over what constitutes legitimate learning versus illicit “distillation” is likely to intensify, potentially evolving into a new battleground of legal and regulatory standards, according to Hsieh.

“AI competition is no longer just about technology,” he said. “It is becoming a contest over rules, law, and intelligence—and ultimately, a matter of national survival.”

Li Jing and Luo Ya contributed to this report.