Anthropic Urges Congress to Crack Down on AI Distillation By Chinese Rivals

Anthropic is calling on Congress to strengthen protections against the distillation of AI models, after claiming that Alibaba-affiliated operators have made the largest known effort to extract capabilities from its 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 Qwen AI lab generated more than 28.8 million exchanges with Claude between April 22 and June 5 using nearly 25,000 “fraudulent accounts,” or accounts that do not represent real, organic users.

Known as a distillation attackAccording to Anthropic, operations focused on Claude’s agentic reasoning, software engineering, and long-horizon planning capabilities, allowing competitors to reproduce advanced model behavior without the expense of training a groundbreaking AI system.

“In addition to its size, this campaign was notable for its brazen nature,” Anthropic wrote. “Alibaba is listed on the New York Stock Exchange, has operations in the United States and is accountable to U.S. investors and regulators.”

Anthropic said the campaign went beyond intellectual property concerns and framed large-scale model distillation as a national security issue that could accelerate China’s military and cyber AI capabilities while reducing the United States’ technological lead.

The letter comes as Washington steps up efforts to protect America’s AI leadership. Earlier this month, President Donald Trump signed an executive order expanding AI-powered cybersecurity initiatives after delaying the measure over concerns it could weaken America’s competitive position vis-à-vis China.

“When PRC labs distill these capabilities from American models, they reap the returns on American investments without incurring the costs or risks associated with training groundbreaking AI models,” Anthropic wrote. “This reverses the economic logic that underpins U.S. AI leadership, turning billions of dollars in research and development, computing, and other U.S. investments into a subsidy for our competitors.”

Anthropic urged lawmakers to expand intelligence sharing between pioneering AI developers and the U.S. government, clarify antitrust rules to allow AI companies to share information about distillation attacks, strengthen export controls on advanced AI chips and computers, close loopholes allowing Chinese companies to access overseas data centers, and impose fines on companies responsible for large-scale model extraction.

A spokesperson for Anthropic declined to comment specifically on the letter, but did share it Declutter“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 America’s leadership in AI.”

The letter also builds on Anthropic’s claims in February that Chinese AI developers DeepSeek, Moonshot AI and MiniMax generated more than 16 million Claude exchanges using approximately 24,000 fraudulent accounts.

These accusations drew criticism from observers who claimed that AI companies rely on similar techniques when training their own systems. Anthropic has countered that conventional distillation is a legitimate method of producing smaller, cheaper models, while the unauthorized extraction of frontier model capabilities through fraudulent access violates its terms of service.

The broader debate over distillation has become more complicated in recent months. In April, Elon Musk testified in federal court that xAI had “partially” used OpenAI models while training Grok, underscoring that distillation is an established industry practice even as companies dispute where legitimate model training ends and unauthorized model extraction begins.