Accessibility help Skip to navigation Skip to main content Skip to footer

Artificial intelligence

Add to myFT

Get instant alerts for this topic

Manage your delivery channels here Remove from myFT

Anthropic accuses Alibaba of obtaining illicit access to Claude

AI company says Chinese ecommerce group used fake accounts to ‘extract’ chatbot’s capabilities

An illustration showing the Claude logo on a screen, with an orange burst design in the background.Anthropic said Alibaba used 25,000 fraudulent accounts to generate over 28.8mn exchanges with Claude© Reuters

current progress 0%

Demetri Sevastopulo in Washington and Cristina Criddle in San Francisco

Publishedyesterday

Updated01:22

Jump to comments section Print this page

Stay informed with free updates

Simply sign up to the Artificial intelligence myFT Digest -- delivered directly to your inbox.

Anthropic has accused Chinese ecommerce giant Alibaba of obtaining illicit access to Claude by creating fake accounts designed to access the AI model which the American company does not offer to Chinese groups.

In a letter to Congress, Anthropic said Alibaba had conducted the “largest campaign to illicitly extract Claude’s capabilities”. It said the Chinese company used 25,000 fraudulent accounts to generate more than 28.8mn exchanges with Claude, which it said was a violation of its terms of service.

“Alibaba’s campaign targeted some of Claude’s most valuable capabilities, such as agentic reasoning, software engineering, and long-horizon tasks,” Anthropic wrote in a June 10 letter to the Senate Banking Committee.

Anthropic urged Congress to “close loopholes allowing PRC [People’s Republic of China] AI labs to access advanced US chips, and penalise PRC labs responsible for distillation attacks”, according to the letter, which was obtained by the FT.

The accusation comes as US pressure mounts on Alibaba, which is involved in businesses from ecommerce to cloud computing. The company this week asked a US court to remove it from a Pentagon blacklist of Chinese companies with alleged ties to China’s People’s Liberation Army.

The FT in November reported the White House had concluded that the Hangzhou-based group had provided technological support for Chinese military “operations” against unspecified US targets. The claim, which the company strongly denied, came in a White House memo obtained by the FT that was based on declassified “top secret” intelligence.

Alibaba has denied having any connection to the PLA. It did not respond to a request for comment about the accusations from Anthropic.

Shares of the Chinese tech group fell more than 4 per cent in Hong Kong on Thursday.

Anthropic has previously accused Chinese AI labs DeepSeek, Moonshot and MiniMax of so-called distillation, claiming they “illicitly” used Claude’s outputs to train their respective models. OpenAI has also accused Chinese groups of engaging in the same behaviour.

In its letter, Anthropic said Alibaba’s campaign to access Claude occurred between late April and early June. It said the distillation attack followed the same pattern as DeepSeek, Moonshot and MiniMax, which had generated a total of 16mn exchanges with Claude through 24,000 fraudulent accounts.

The AI company said Alibaba had “brazenly” run the campaign after the White House issued a memo on the need to curb the industrial-scale theft of AI labs’ intellectual property, which was first reported by the FT.

A White House official said “industrial distillation” of US AI models was “unacceptable”.

“We intend to hold bad actors accountable,” the official added.

Distillation is used widely across the technology industry to train cheaper and smaller versions of models from companies. However, American AI companies have argued that if Chinese competitors use it to acquire advanced capabilities, it could have serious national security repercussions.

Security fears are particularly acute as they pertain to Anthropic’s Mythos model, which has advanced cyber capabilities. The US government had been working with Anthropic to apply the model in cyber-offensive operations against China and other adversaries.

The FT reported in September that the San Francisco-based Anthropic had become the first American AI company to stop selling artificial intelligence services to groups majority owned by Chinese entities. An Anthropic executive said it was trying to limit the ability of Beijing to use its technology to benefit China’s military and intelligence services.

Anthropic’s letter to Congress was first reported by Bloomberg.

Reuse this content(opens in new window) CommentsJump to comments section

Follow the topics in this article

Add to myFT

Add to myFT

Add to myFT

Add to myFT

Add to myFT

Comments

Close side navigation menu

Search the FTSearch

Subscribe for full access

Read Original at Financial Times →