Anthropic Names Alibaba, DeepSeek in AI Distillation Report
Anthropic said the campaigns used hundreds of millions of Claude exchanges to train rival models and extract reasoning traces without authorization.
- On Thursday, Anthropic released a report alleging seven China-based AI developers, including Alibaba, DeepSeek, Moonshot, and Xiaomi, conducted large-scale campaigns to extract Claude's capabilities and replicate them in competing products.
- The report describes 'illicit distillation,' where unauthorized labs use fraudulent accounts to harvest Claude outputs without authorization. Anthropic observed nearly 200 million exchanges linked to five separate distillation campaigns conducted over recent months.
- Alibaba operated the largest campaign, generating more than 151 million exchanges between May and July. Moonshot and DeepSeek silently routed live customer conversations through Claude without user knowledge to train their own models.
- Some exchanges contained sensitive credentials, including data from a Russian government database and a Chinese police system. Anthropic banned associated accounts, though Alibaba, Moonshot, and DeepSeek did not respond to requests for comment.
- Technical forum commenters questioned whether Anthropic can accurately distinguish Chinese AI labs from third-party token resellers operating on the same infrastructure. OpenAI and Google have previously published reports attributing similar distillation activities to DeepSeek.
14 Articles
14 Articles
Günstige KI aus China weckt Zweifel an den Marktchancen großer US-Anbieter wie Anthropic. Das Unternehmen bericht nun von dreisten Tricksereien chinesischer
Antropic has revealed evidence that Chinese artificial intelligence (AI) companies used thousands of fake accounts and overseas intermediary networks to collect large volumes of responses from its AI model, 'Claude,' and attempted to utilize them for the development of its own models. According to a threat intelligence report released by Antropic on the 10th (local time), Chinese AI companies such as Moonshot AI and DeepSeek used so-called 'tran…
Chinese AI labs secretly used millions of Claude exchanges to train their models, Anthropic says
Anthropic said it detected unauthorized efforts by China-based AI labs including Alibaba and Moonshot AI to use its Claude models to help improve their own AI systems.
(Seoul = Yonhap News) Reporter Jung Joo-ho = Antropics exposed the detailed distillation methods used by Chinese artificial intelligence (AI) companies utilizing fake accounts, and its model 'Claude...
Coverage Details
Bias Distribution
- 43% of the sources are Center, 43% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
