Media365
US artificial intelligence company Anthropic has accused Chinese technology giant Alibaba of carrying out what it described as the largest known attempt to illicitly extract capabilities from its Claude AI models, Reuters reported, escalating tensions over intellectual property theft and the intensifying technological rivalry between the United States and China.
According to the report, Anthropic alleged that operators affiliated with Alibaba and its AI research arm Qwen conducted a large-scale “distillation” campaign between April 22 and June 5 this year. The effort allegedly generated more than 28.8 million interactions with Anthropic’s Claude platform through nearly 25,000 fraudulent accounts.
Anthropic warned that Alibaba and other Chinese labs are systematically and unauthorisedly using outputs from leading US models to develop rival chatbots at a fraction of the cost through a practice it described as adversarial distillation. The company said AI systems built using this method often lack safety guardrails.
Distillation refers to a technique in which a smaller or less advanced AI model is trained using the outputs of a more powerful model. While the process is commonly used within companies to improve efficiency, AI developers have increasingly accused rivals of using it to replicate proprietary capabilities without authorisation.
Anthropic said the campaign was aimed at accelerating China’s ability to match the capabilities of its advanced Mythos Preview model, one of the company’s latest frontier AI systems.
Letter sent to US lawmakers
The allegations were detailed in a June 10 letter sent by Anthropic to US Senators Tim Scott and Elizabeth Warren, the chair and ranking member of the Senate Banking Committee, ahead of a hearing focused on artificial intelligence and national security.
In the letter, Anthropic said the alleged campaign was conducted “illicitly, systematically, and at an industrial scale” to harvest US AI capabilities across frontier labs and repackage them as Chinese-owned products without bearing the training and research and development costs required to build frontier models.
The company said the campaign took place after White House Office of Science and Technology Policy director Michael Kratsios published an April memo warning that the US would help crack down on attempts by Chinese companies to exploit outputs from American AI models. The memo described such activity as distinct from legitimate research because of its “industrial scale” and reliance on thousands of proxy accounts.
Anthropic said the alleged Alibaba campaign was carried out in defiance of those warnings and cautioned that failing to respond could allow China to gain ground on the US in AI, posing a threat to national security.
Not the first accusation against Chinese AI firms
The latest allegations follow similar claims made by Anthropic earlier this year.
In a February blog post, the company said it had identified attempts by Chinese AI startup DeepSeek and other Chinese AI developers to extract capabilities from Claude. Anthropic claimed that DeepSeek’s operation involved more than 150,000 interactions with its platform.
The company also alleged that Moonshot AI generated more than 3.4 million exchanges, while MiniMax accounted for over 13 million interactions.
At the time, Anthropic warned that such campaigns were becoming increasingly sophisticated and frequent, arguing that tackling the problem would require coordinated action from governments, technology companies and the broader AI ecosystem.
The practice has alarmed US developers to the point that Anthropic, OpenAI and Alphabet’s Google have joined forces to share information about distillation attempts that violate their terms of service. Anthropic and OpenAI have each warned that Chinese AI start-ups, including DeepSeek and MiniMax, have used distillation to develop their own models.
AI race increasingly shaped by geopolitics
The allegations arrive at a time when AI development has become deeply intertwined with national security concerns.
Earlier this month, Alibaba was added to the Pentagon’s list of Chinese military-linked companies, a designation the company is challenging. Inclusion on the list does not automatically trigger sanctions but can complicate business relationships and heighten regulatory scrutiny.
Alibaba has insisted it has no affiliation with the Chinese military and sued the Pentagon this week to seek removal of the designation.
At the same time, the US government has adopted a mixed approach towards Chinese AI firms. Reuters recently reported that the Commerce Department has not yet placed DeepSeek on a trade blacklist despite national security concerns raised by an interagency review panel, partly to avoid further escalating tensions with Beijing.
Lawmakers in Washington are now moving to address the industry’s concerns. In the Senate, Tennessee Republican Bill Hagerty and New Jersey Democrat Andy Kim plan to introduce an amendment to must-pass defence legislation as soon as Wednesday that would blacklist or impose sanctions on any Chinese firm found to be improperly accessing US AI model output to help train competing models, Reuters reported.
Anthropic under pressure too
Anthropic itself has also come under tighter government oversight.
On June 12, just two days after the company sent its letter to lawmakers, the US Commerce Department imposed restrictions on Anthropic’s latest Mythos and Fable AI models over concerns that they could be used by military or intelligence entities in China and other countries deemed sensitive by Washington. The restrictions prompted Anthropic to disable access to the models globally.
The move highlighted the increasingly complex balance between commercial AI innovation, national security considerations and international competition.
For Anthropic, the threat of cheaper imitation products from China that could siphon away customers looms large as the company, now valued by private investors at $US965 billion, prepares for a potential initial public offering.
At the same time, Anthropic’s calls for stronger government action may not find a fully receptive audience in Washington. The company is embroiled in a fresh dispute with the Trump administration over the export controls imposed on its top two models, underscoring how even US AI leaders are increasingly caught in the crossfire of the global technology race.
Why this matters
The dispute underscores a growing challenge facing leading AI companies: protecting advanced models from rivals seeking to replicate their capabilities.
As AI systems become more powerful and economically valuable, accusations of model distillation and intellectual property theft are emerging as a new front in the technology competition between the United States and China.
For investors and policymakers alike, the case signals that future battles in the AI race may be fought not only through breakthroughs in computing power and algorithms but also through access controls, cybersecurity measures and regulatory action.
