- Law4Startups
- Posts
- ⚖️ Anthropic and Chinese AI Labs
⚖️ Anthropic and Chinese AI Labs
Anthropic vs. Chinese AI Labs
Anthropic accused three of China’s most prominent AI laboratories—DeepSeek, Moonshot AI, and MiniMax—of orchestrating a massive, "industrial-scale" campaign to siphon the intelligence of its Claude models. The accusation centers on distillation, a technique where a smaller or newer model is trained using the high-quality outputs of a superior model. While distillation is a standard industry practice for optimization, Anthropic argues that these labs used 24,000 fraudulent accounts to generate over 16 million exchanges, effectively "copying the homework" of Claude to bypass years of R&D and millions in safety training costs.
"Extraction" Campaigns and Fallout
Anthropic’s forensic report suggests that each lab targeted specific high-value capabilities that distinguish Claude in the marketplace. By using "Hydra Cluster" architectures—networks of fake accounts designed to evade rate limits—the labs reportedly captured nuanced reasoning traces that are typically hidden from public view.
The timing of these accusations is not accidental. It coincides with an intense debate in Washington over the Trump administration's recent decision to allow exports of the NVIDIA H200 chip to China. Anthropic is leveraging this "distillation" evidence to lobby for stricter controls, arguing that the scale of these attacks requires high-end compute to process the siphoned data.
Furthermore, Anthropic claims these distilled models pose a unique security risk. While Claude is trained with "Constitutional AI" guardrails to refuse requests for bioweapon formulas or malicious code, a distilled model captures the capabilities without the safeguards. This allows foreign entities to effectively "strip-mine" American AI for offensive cyber operations while discarding the safety filters Anthropic spent billions to install.
In addition to our newsletter we offer 60+ free legal templates for companies in the UK, Canada and the US. These include employment contracts, investment agreements and more
