US artificial intelligence company Anthropic said on Monday it had uncovered campaigns by three Chinese AI firms to illegally extract capabilities from its cloud chatbots, in what it said was intellectual property theft on an industrial scale. OpenAI made similar allegations last month.
Anthropic said DeepSeek, Moonshot AI and MiniMax used a technique called “distillation” – using output from more powerful AI systems to rapidly increase the performance of less capable systems.
“These campaigns are increasing in intensity and sophistication,” the company said in a statement. “The window to act is narrow.”
Distillation is a common practice within AI development, often used by companies to create cheaper, smaller versions of their own models.
This practice made headlines a year ago, when the release of DeepSeek’s low-cost generative AI model performed at the same level as ChatGPT and other top US chatbots, overturning perceptions of US dominance in the sensitive area.
Anthropic said the companies achieved their target through their cloud model and about 16 million exchanges with 24,000 fake accounts.
This allowed the three labs to snatch up capabilities they had not developed independently at a fraction of the cost – and in doing so circumvent export controls on powerful American technology aimed at preserving American dominance in AI.
The company argued that the practice threatens national security, saying that models made through illegal distillation are unlikely to maintain safeguards designed to prevent abuse – such as bans on helping develop bioweapons or enabling cyberattacks.
Anthropic’s arch rival OpenAI, creator of ChatGPT, made similar allegations to US lawmakers earlier this month, saying Chinese companies were using the technology amid “ongoing efforts to free ride on capabilities developed by OpenAI and other US frontline laboratories”.
Anthropic said Minimax ran the largest operation, generating more than 13 million exchanges. Each campaign focuses heavily on coding, agentic reasoning, and tool usage – areas where the cloud is considered a leader.
To circumvent Anthropic’s ban on commercial access from China, the labs allegedly routed traffic through proxy services that managed vast networks of fraudulent accounts.
Anthropic called for coordinated industry and government responses saying no one company can tackle the situation alone.
