Anthropic detected and disrupted unauthorized use of its Claude models by Chinese AI labs, including Alibaba, Moonshot, and DeepSeek, involving over 151 million exchanges. The U.S. company described the activity as 'illicit distillation' in a threat intelligence report released Thursday.
TL;DR
- Anthropic uncovered large-scale unauthorized use of Claude by Chinese AI labs to train their models.
- Alibaba, Moonshot, and DeepSeek were among the companies involved, with Alibaba's campaign peaking at nearly 3 million exchanges per day.
- The activity raises concerns about privacy laws and the labs' terms of service compliance.
What happened
Anthropic's report detailed how Alibaba used Claude outputs to train its Qwen models, involving over 151 million exchanges between May and July. The activity peaked at nearly 3 million exchanges per day from more than 3,500 fraudulent accounts, according to the report.
Moonshot AI silently forwarded customer requests intended for its Kimi models to Claude, using the responses to train its own models. In one 10-day period, Moonshot relayed nearly 300,000 customer requests to Claude, with the majority routed to Claude Opus models. Moonshot saved some of these exchanges and extracted Claude's reasoning transcripts for training data, according to the report.
DeepSeek also transferred exchanges to Claude without notifying its customers. Anthropic observed more than 12 million distillation attacks attributable to DeepSeek over 14 days in July 2026. The report covers activity disrupted between December 2025 and August 2026 across seven areas, including cyber operations, influence operations, and biological misuse.
Why it matters
This incident highlights the risks of unauthorized model training and the potential misuse of sensitive information. The activity raises concerns about compliance with privacy laws and the labs' own terms of service, according to the report.
For AI developers and startups, this underscores the importance of robust security measures and monitoring to prevent unauthorized use of AI models. It also highlights the competitive pressures in the AI industry and the lengths some companies may go to gain an edge.
Investors should be aware of the potential risks and ethical considerations associated with AI model training and usage. This incident may also impact the reputation and trustworthiness of the companies involved, which could have implications for their market position and valuation.
Key facts
- Anthropic detected unauthorized use of its Claude models by Chinese AI labs, including Alibaba, Moonshot, and DeepSeek.
- Alibaba's campaign involved over 151 million exchanges with Claude between May and July, peaking at nearly 3 million exchanges per day.
- Moonshot relayed nearly 300,000 customer requests to Claude in a 10-day period, using the responses to train its own models.
- DeepSeek transferred over 12 million exchanges to Claude without notifying its customers over 14 days in July 2026.
- The activity involved sensitive information from individual users, major multinational companies, and state-affiliated actors.
- Anthropic described the activity as 'illicit distillation', a process where outputs from a more capable AI model are used to train another model without authorization.
- The report covers activity disrupted between December 2025 and August 2026 across seven areas, including cyber operations, influence operations, and biological misuse.
- Alibaba, Moonshot, DeepSeek, Xiaomi, and Anthropic did not immediately respond to CNBC's requests for comment.
Context
This incident is part of a broader trend of AI model misuse and unauthorized training. As AI models become more powerful and widely used, the risk of such activities increases, highlighting the need for robust security measures and ethical guidelines in the AI industry.
The competitive landscape in the AI industry is intense, with companies constantly seeking to gain an edge through innovative models and training techniques. However, this incident underscores the importance of ethical considerations and compliance with privacy laws in the pursuit of technological advancement.
For investors, this incident serves as a reminder of the potential risks and ethical considerations associated with AI model training and usage. It also highlights the importance of due diligence and ongoing monitoring of AI companies' practices and reputations.
