Anthropic has identified and disrupted extensive unauthorized activities by Chinese AI labs that involved using outputs from its Claude model to enhance their own AI systems. The report highlights that Alibaba's operation was particularly large, with over 151 million exchanges occurring between May and July, peaking at nearly 3 million exchanges per day through fraudulent accounts.
Moonshot and DeepSeek also engaged in similar practices, with Moonshot relaying nearly 300,000 customer requests to Claude in a short period, often without notifying users. These actions not only breach privacy laws but also violate the terms of service of the involved companies.
The implications of these findings are significant, as they raise questions about the integrity of AI development practices in China and could lead to increased scrutiny and regulatory actions against these firms.
The report also mentions that these activities were part of a broader range of cyber operations and influence tactics observed by Anthropic, indicating a serious challenge for the AI industry regarding data security and ethical standards