AI startups seek additional government support to counter unfair practices by Chinese rivals
issued Thursday, June 25, 2026 · 07:43 AM
ANTHROPIC has accused Chinese tech giant Alibaba of waging a massive effort to “unauthorizedly” access its Claude artificial intelligence models using thousands of fraudulent accounts, undermining a U.S. AI developer’s decision to keep its products out of China.
Anthropic claimed that a campaign by operators associated with Alibaba’s Qwen AI Lab targeted Claude’s most valuable competencies, including software engineering and agent inference, according to a letter the AI startup sent to several U.S. senators and White House officials. The company said this is the largest attempt yet by a Chinese company to piggyback on the research of a top US institute.
Anthropic claimed in the letter that the effort included 28.8 million interactions with Claude through about 25,000 fraudulent accounts between April and June, according to people familiar with the document and a copy obtained by Bloomberg News. The company said Alibaba’s campaign is similar to past efforts by other Chinese developers that Anthropic flagged in a blog post earlier this year.
Anthropic warned that Alibaba and other Chinese labs are systematically exploiting the results of leading US models to develop rival generations of chatbots at a fraction of the cost, using a technique known as adversarial distillation. The company warned that AI systems built using this technique often lack safety guardrails and called on the Trump administration to step up efforts to stop it.
“These distillation attacks are being carried out illegally, systematically, and on an industrial scale to harvest U.S. Al capabilities across Frontier laboratories and repackage them as their own without the training and R&D costs necessary to train the U.S. Frontier model,” Anthropic said in the letter.
Alibaba has not commented. An Anthropic spokesperson did not go into details of the letter, but stressed the importance of combating distillation through “coordinated government and industry action”.
Anthropic’s letter follows recent calls by America’s top AI companies to curb the kind of distillation in which developers use the results of another AI model to train their systems and create similar functionality in new AI models at a much lower cost. While distillation is acceptable for training small, less sophisticated systems, it violates the AI Lab’s Terms of Use when used to replicate cutting-edge AI models without permission.
The practice has alarmed U.S. developers, who say Anthropic, OpenAI, and Alphabet’s Google are partnering to share information about distillation attempts that violate their terms of service. Anthropic and OpenAI each warned that Chinese AI startups, including DeepSeek and Minimax, are using distillation to develop their own models.
Lawmakers in Washington are working to address U.S. industry concerns. Tennessee Republican Bill Hagerty and New Jersey Democrat Andy Kim are expected to introduce amendments as early as Wednesday to the defense bill that would require passage of blacklisting or sanctions against Chinese companies found to have unauthorized access to U.S. AI model output to help train competing models, according to people familiar with the matter.
Related items
It is unclear whether this amendment will gain enough support to be included in the final version of the national defense bill. A companion bipartisan bill in the House, sponsored by Michigan Republican Bill Huizenga and Democrat Sidney Kamlager Dove, will also be considered for inclusion in the annual defense package.
These proposals are in line with the Trump administration’s early steps on the issue. Michael Crasios, director of the White House Office of Science and Technology Policy, released a memo in April suggesting the United States would cooperate in cracking down on attempts by Chinese companies to exploit the work of American models. The memo explains that this differs from legitimate research practice due to its “industrial scale” and reliance on thousands of proxy accounts.
Anthropic said the Alibaba campaign came after Kratsios ignored the administration’s warnings and released the memo. He warned that if China fails to respond to such efforts, there is a risk that China will gain an advantage over the United States in the field of AI, posing a threat to national security.
The allegations against Alibaba have added to the political pressure in Washington on Alibaba, which earlier this month was added to the U.S. Department of Defense’s blacklist of companies that allegedly support the Chinese military (a development cited in Antropic’s letter). Alibaba claims it has no ties to the Chinese military and sued the Pentagon this week seeking to have the designation lifted.
For Anthropic, the threat of cheap Chinese counterfeit products siphoning off customers looms large as the company, currently valued at $965 billion by private investors, prepares for an initial public offering. According to a report by Bloomberg, U.S. authorities estimate that unlicensed distillation is costing Silicon Valley laboratories billions of dollars.
In its letter, Anthropic is asking the U.S. to clarify its antitrust guidelines to allow U.S. companies to better share information about distilling. It reiterated its support for export controls on advanced AI chips and called on the US to penalize companies that use distillation to gather valuable information to create their own models.
Anthropic is seeking additional government support to counter what it sees as unfair practices by its Chinese rivals, but that may not be well received by the White House. The company is embroiled in a new dispute with the Trump administration. Less than two weeks ago, the Trump administration imposed export restrictions on two of Anthropic’s top models, citing security concerns.
Even after a meeting between Anthropic’s top technical staff and White House officials last week, little progress has been made to ease tensions and restore service to the company’s Fable 5 and Mythos 5 AI systems. The company disabled access to its models more than a week ago after the Commerce Department imposed restrictions blocking foreigners from using these AI tools. bloomberg
