US AI giants accuse Chinese rivals of massive data theft

San Francisco., The American artificial intelligence (AI) company Anthropic yesterday claimed to have discovered campaigns by three Chinese AI firms to illicitly extract capabilities from its Claude chatbot, in what it described as intellectual property theft on an industrial scale.

Anthropic noted that DeepSeek, Moonshot AI and MiniMax used a technique known as distillation, which involves using the results of a more powerful AI system to quickly improve the performance of a less capable one.

“These campaigns are growing in intensity and sophistication,” the company said in a statement. “The window for action is narrow.”

Distillation is a common practice in AI development, often used by companies to create cheaper, smaller versions of their own models.

The practice made headlines a year ago, when DeepSeek’s launch of a low-cost generative AI model performed on par with ChatGPT and other leading US chatbots, upending assumptions about US hegemony in the sector.

Anthropic claimed that the companies achieved their goals through approximately 16 million exchanges with its Claude model and 24,000 fake accounts.

This allowed the three labs to extract capabilities they had not developed independently, at a fraction of the cost. In doing so, they circumvented export controls on advanced US technology, the company added.

Anthropic argued that this practice poses risks to national security, as models built through illicit distillation are unlikely to maintain security mechanisms designed to prevent misuse, such as restrictions to help develop biological weapons or facilitate cyberattacks.

OpenAI, Anthropic’s archrival and creator of ChatGPT, brought similar allegations to US lawmakers earlier this month.

Anthropic called for coordinated responses from industry and government to address what it says no company can address alone.