Chinese AI companies accused of stealing US tech through distillation, threatening national security

US AI Labs Accuse Chinese AI Companies of Systematic Technology Theft

Leading American artificial intelligence companies are raising urgent concerns that Chinese AI companies are systematically copying their most advanced technologies through a technique called distillation. On June 24, Anthropic sent a letter to the Senate Banking Committee directly accusing Alibaba of conducting AI model theft through what it described as "brazen" distillation attacks 2

2

. The San Francisco-based company alleged that Alibaba employees illicitly generated 28.8 million outputs from its Claude chatbot using nearly 25,000 fraudulent accounts between April 22 and June 5 2

2

.

Sam Altman's OpenAI has similarly accused DeepSeek of "ongoing efforts to free-ride" on its work, while Google warned of surging malicious activity originating from China, Iran, North Korea and Russia 2

2

. These accusations come as experts estimate China now trails the United States in AI development by just six months 1

1

. The Chinese startup Z.ai recently released GLM-5.2, an AI model nearly as powerful as top American systems, particularly in cybersecurity applications that the Trump administration considers vital to geopolitical competition 1

1

.

How Distillation Enables Chinese Copycats to Bypass Development Costs

Distillation itself is not new—Google researchers first developed the technique in the early 2010s as a legitimate method for building more efficient AI systems 1

1

. The process allows researchers to collect data from a powerful system and use it to train a less capable model that can run on cheaper hardware. "Think of one model as the teacher and the other as a student," explained Geoffrey Hinton, a former Google researcher who helped develop the technique 1

1

.

The controversy arises when companies use distillation to mimic proprietary AI technologies built by competitors without permission. Both Anthropic and OpenAI explicitly prohibit distillation of their leading systems under their terms of service, yet the practice remains widespread 1

1

. During April testimony in a federal trial in Oakland, California, even Elon Musk acknowledged the practice at his AI company xAI, stating "Generally AI companies distill other AI companies" 1

1

.

For China, distillation serves as a crucial strategy to circumvent US export controls on powerful computer chips needed to build advanced models, such as those sold by Nvidia 2

2

. "They effectively are bypassing United States chip bans by using our own multi-billion-dollar AI model to build their own," said Theresa Payton, who served as White House chief information officer under President George W. Bush 2

2

.

Scale of Illicitly Harvesting Data Raises US National Security Concerns

The scale of alleged data harvesting has alarmed both AI companies and government officials. In February, Anthropic accused DeepSeek and two other Chinese startups of improperly harvesting large amounts of data from its systems, claiming the startups used about 24,000 accounts to generate more than 16 million conversations with its Claude chatbot 1

1

. These industrial-scale theft campaigns allow Chinese firms to build competitive models at rock-bottom prices, helping them steal US business customers by offering cheaper "tokens" needed to use AI services 2

2

.

Six of the top 10 most-used AI models are now developed by Chinese tech firms, including Z.AI, Minimax and DeepSeek—all boasting capabilities nearly as advanced as Anthropic's Claude or OpenAI's ChatGPT 2

2

. A source close to Anthropic estimated that without distillation, China would be roughly 18-plus months behind in frontier models rather than the current 6-9 months 2

2

.

Anthropic has regularly briefed Trump administration national security officials about distillation attacks, including as recently as this month. "We have basically spoken to every part of the government about this," including the White House and Congress, sources told The Post 2

2

. The company's powerful "Mythos" and "Fable" models were temporarily taken offline due to White House concerns they could be weaponized by China 2

2

.

Legal Ambiguity and Geopolitical Implications of Ripping Off Cutting-Edge AI

The legal status of distillation remains murky. Sarah Tishler, a partner at law firm Beck Reed Riden specializing in trade-secret litigation, notes that while some legal scholars argue the practice violates the 2016 Defend Trade Secrets Act, courts have not explicitly decided the matter 1

1

. Copyright law may not apply because distillation copies system behavior rather than text verbatim 1

1

.

The Trump administration has identified distillation as a key challenge in the race to develop advanced AI. In April, the White House Office of Science and Technology published a memo accusing "foreign entities, principally based in China" of engaging in "deliberate, industrial-scale" theft campaigns 2

2

. OpenAI briefed the Congressional-Executive Commission on China in June on China-linked misuse of AI, including influence operations, and advocated for the NSA's AI Security Center to serve as a centralized hub for sharing details on distillation attacks 2

2

.

Experts warn that unchecked distillation could erode US leadership in AI and enable adversaries to develop terrifying hacking or autonomous weapon capabilities. The stakes extend beyond commercial competition to fundamental questions of national security and technological supremacy. As Payton characterized it, China's distillation attacks represent the "ultimate form of industrial homework theft," allowing foreign adversaries to "clone the smartest kid in the class for pennies on the dollar" 2

2

. The coming months will test whether US policymakers can develop effective countermeasures to protect proprietary AI technologies while maintaining America's technological edge.