Anthropic Says Chinese Labs Used Millions of Claude Exchanges to Train Models
Anthropic has accused Chinese AI labs of secretly using millions of Claude exchanges to train rival models, framing the activity as large-scale illicit distillation rather than routine API use. Reporting tied to the Wall Street Journal and Techmeme names DeepSeek and Moonshot, alleging thousands of fake accounts and millions of real user queries routed through overseas transfer stations outside China. Bloomberg separately reports Moonshot covertly sent user requests to Claude and repackaged the outputs as its own. Reuters adds that Anthropic disrupted Russian and Chinese campaigns targeting Claude, and the company published a broader threat intelligence report covering cyber operations, influence work, surveillance, and biological misuse. The disclosures mark a shift from terms-of-service enforcement toward active counter-distillation monitoring, though independent verification of training use and detection methods remains limited in the public record.
Anthropic Says Chinese Labs Used Millions of Claude Exchanges to Train Models
CNBC reports Anthropic alleges Chinese AI labs secretly used millions of Claude exchanges to train their models. Reuters separately reports Anthropic disrupted Russian and Chinese AI campaigns targeting Claude.
Key takeaway
Frontier API providers now face industrial-scale distillation attempts that bypass contractual limits, forcing technical enforcement beyond trust-based access controls.
What happened
CNBC reports Anthropic alleges Chinese AI labs secretly used millions of Claude exchanges to train their models, describing an unauthorized distillation effort at scale rather than isolated misuse.
Wall Street Journal reporting summarized by Techmeme says DeepSeek and Moonshot used thousands of fake accounts and millions of real user queries sent via transfer stations outside China, while Bloomberg reports Moonshot routed user requests through Claude and passed off responses as its own.
Evidence
Anthropic alleges Chinese AI labs secretly used millions of Claude exchanges to train their models.
CNBC AI · attributed
CNBC reports Anthropic alleges Chinese AI labs secretly used millions of Claude exchanges to train their models.
DeepSeek and Moonshot used thousands of fake accounts and millions of real user queries in a distillation effort routed through transfer stations outside China.
Techmeme · attributed
Anthropic says DeepSeek and Moonshot used thousands of fake accounts and millions of real user queries in 'distillation' effort to clone its AI capabilities
Moonshot covertly routed user requests to Claude and passed off the responses as its own.
Bloomberg Technology · attributed
Anthropic PBC accused China's artificial intelligence champion Moonshot AI of covertly routing thousands of user requests to the US firm's Claude models and passing off the responses as its own in a bid to gain an edge in the AI race.
Anthropic disrupted Russian and Chinese AI campaigns targeting its Claude models.
Reuters AI · attributed
Reuters separately reports Anthropic disrupted Russian and Chinese AI campaigns targeting Claude.
Anthropic published a threat intelligence report on disrupted misuse spanning cyberattacks, influence operations, surveillance, and illicit distillation.
Techmeme · attributed
Anthropic publishes a threat intelligence report on how it disrupted efforts to misuse Claude for cyberattacks, influence operations, surveillance, and more
Why it matters
The episode shows closed-model providers treating API traffic as an active intelligence and enforcement surface amid geopolitical competition over frontier AI capabilities.
Limits and uncertainties
Public excerpts do not specify how Anthropic detected the activity or independently verify that scraped exchanges were used in rival model training.
Bloomberg and Techmeme accounts rest on Anthropic allegations without disclosed technical proof in the packet.
Practical implications
API operators should assume high-volume query patterns may be monitored for distillation signatures and plan rate limits, account integrity checks, and output controls accordingly.
Builders relying on third-party model APIs need tighter logging and anomaly detection because proxy routing via overseas transfer stations can mask origin traffic.
What to watch
Whether Anthropic, regulators, or named firms publish technical attribution details, enforcement actions, or responses from DeepSeek and Moonshot.
Follow-on reporting on additional disrupted Russian and Chinese campaigns described in the threat intelligence release.
