Anthropic Accuses Chinese AI Firms of Data Theft: A Deeper Look

In a recent report, Anthropic, a prominent American AI unicorn, has accused three leading Chinese AI companies—DeepSeek, Moonshot AI, and MiniMax—of executing large-scale "distillation attacks" on its Claude model. These attacks allegedly involved over 16 million conversation exchanges through approximately 24,000 fraudulent accounts, aimed at stealing Claude's core capabilities to train competing models.

What Are Distillation Attacks?

Unlike traditional hacking, these attacks exploit the API's intended functionality but for unintended purposes. The Chinese firms reportedly used a "Hydra" architecture, dispersing requests through vast proxy networks and blending them with mundane queries to avoid detection. Each company focused on different aspects: MiniMax targeted agentic coding, Moonshot AI focused on agentic reasoning and computer vision, and DeepSeek specialized in extracting "chain-of-thought" data, even manipulating Claude to rewrite politically sensitive content.

Implications and Risks

Anthropic warns that such industrial-scale extraction not only harms commercial interests but also poses national security risks. Distillation can strip away safety guardrails, making resulting models prone to misuse. Furthermore, this practice threatens the AI industry's economic model, where billions are invested in R&D, only for competitors to replicate capabilities cheaply via API calls.

Counterarguments and Hypocrisy Claims

While Anthropic's accusations are severe, some analysts argue that Chinese firms like Moonshot AI, with its Kimi K2 model and proprietary MuonClip optimizer, demonstrate significant independent innovation. Critics also point out a perceived hypocrisy, noting that Anthropic itself has faced lawsuits for data scraping. For instance, Reddit sued Anthropic in 2025 for allegedly scraping user comments to train Claude without permission. Additionally, Anthropic settled a $1.5 billion lawsuit with authors over the use of pirated material for chatbot training.

Elon Musk's Perspective

Adding fuel to the fire, Elon Musk, founder of xAI, commented on the situation via a retweet, criticizing Anthropic with a remark akin to "the pot calling the kettle black," highlighting the irony given Anthropic's own legal battles over data usage.

Conclusion

The controversy surrounding Anthropic's accusations against Chinese AI firms raises critical questions about ethics, innovation, and data usage in the AI industry. As legal battles and criticisms mount on all sides, the need for clearer regulations and international cooperation becomes evident. What are your thoughts on this complex issue? Let us know in the comments!