Anthropic urges Congress to crack down on AI distillation by Chinese competitors

Anthropic is calling on Congress to strengthen protections against the distillation of AI models after it alleged that Alibaba’s operators conducted the largest known effort to extract capabilities from their Claude chatbot.

In a June 10 letter to Senate Banking, Housing, and Urban Affairs Committee Chairman Tim Scott and Ranking Member Elizabeth Warren, Anthropic alleged that operators affiliated with Alibaba and its Cowen AI Lab made more than 28.8 million exchanges with Cloud between April 22 and June 5 using nearly 25,000 “fraudulent accounts,” or those that do not represent real, organic users.

Known as A Distillation attackAnthropic said the operations targeted Claude’s logical reasoning, software engineering, and long-range planning capabilities, allowing competitors to reproduce advanced model behavior without the cost of training a frontier AI system.

“Beyond its scope, this campaign was striking for its brazen nature,” Anthropic wrote. “Alibaba is listed on the New York Stock Exchange, maintains business operations in the United States, and is accountable to US investors and regulators.”

Anthropic said the campaign went beyond intellectual property concerns, framing the broad distillation model as a national security issue that could accelerate China’s military and cyber AI capabilities while narrowing the United States’ technological lead.

The letter comes at a time when Washington is intensifying its efforts to protect American leadership in the field of artificial intelligence. Earlier this month, President Donald Trump I fell Executive order expands AI-powered cybersecurity initiatives yet delay This measure comes due to concerns that it may weaken America’s competitive position against China.

“When PRC labs extract these capabilities from US models, they receive returns on US investments without bearing the costs or risks associated with training frontier AI models,” Anthropic wrote. “This reflects the economic logic that ensures American leadership in artificial intelligence, and turns billions of dollars worth of American research and development, computing, and other investments into support for our competitors.”

Anthropic urged lawmakers to expand intelligence sharing between frontier AI developers and the US government, clarify antitrust rules to allow AI companies to share information about distillation attacks, strengthen export controls on advanced AI and computing chips, close loopholes that allow Chinese companies to access offshore data centers, and impose sanctions on companies responsible for large-scale model mining.

An Anthropic spokesperson declined to comment specifically on the letter, but said so Decryption“We believe that combating the threat of illicit distillation requires coordinated action between government and industry, and we will continue to work with Congress and the administration to maintain American leadership in artificial intelligence,” he added.

The message also relies on anthropomorphism Claims In February, Chinese AI developers DeepSeek, Moonshot AI and MiniMax generated more than 16 million Claude exchanges using nearly 24,000 fraudulent accounts.

These claims sparked criticism from observers who said that AI companies rely on similar techniques when training their own systems. Anthropic has responded that traditional distillation is a legitimate means of producing smaller, cheaper models, while unauthorized extraction of frontier models’ capabilities through fraudulent access violates its terms of service.

The broader debate over distillation has become more complex in recent months. In April, Elon Musk to attest In federal court, xAI used OpenAI models “partially” while training Grok, confirming that distillation is an established industry practice — even as the companies dispute where legitimate model training ends and unauthorized model mining begins.