Technology
Anthropic alleges Chinese AI firms ran large-scale distillation campaigns, report says
Anthropic reported massive data-extraction efforts by Chinese developers targeting Claude models, according to TechCrunch.
The short version
- Anthropic alleged that Chinese artificial intelligence firms targeted its systems in repeated distillation efforts to capture advanced model capabilities.[TechCrunch]
- The company observed nearly 200 million exchanges across five separate extraction campaigns.[TechCrunch]
- Anthropic attributed 151 million exchanges to Alibaba and nearly 300,000 requests over 10 days to Moonshot AI.[TechCrunch]
Key facts
- Anthropic reported that artificial intelligence companies based in China mounted persistent distillation campaigns against US frontier models.[TechCrunch]
- Anthropic recorded roughly 200 million total exchanges tied to distillation efforts across five distinct campaigns.[TechCrunch]
- Alibaba was linked to the largest share of attempts, comprising 151 million interactions between May and July 2026.[TechCrunch]
- A separate campaign linked to Moonshot AI generated nearly 300,000 requests focused on the Opus model across a 10-day span.[TechCrunch]
What remains uncertain
- The findings rely solely on Anthropic's internal attribution and telemetry, without public responses or verification from Alibaba, Moonshot AI, or independent third parties.[TechCrunch]
Sources
Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.