Anthropic Identifies Over 23 Million Hidden Claude Exchanges by Moonshot AI
Anthropic alleges Beijing-based competitor Moonshot routed nearly 300,000 requests and more than 23 million interactions through its Claude models without user consent. A separate report highlights that Russian, Chinese, and Malian operators employed Claude to automate cyberattacks and build state surveillance systems.
Anthropic accused Moonshot AI of secretly sending customer requests intended for its Kimi chatbot to Anthropic's Claude models. The US company alleged that the Beijing-based competitor concealed this practice from users. During a specific ten-day period identified by Anthropic, Moonshot routed almost 300,000 requests through a network of 5,380 fraudulent accounts, with significant traffic directed toward Claude Opus. Between May and July, Anthropic attributed more than 23 million Claude exchanges to Moonshot. The company stated it used intermediaries and large numbers of accounts to evade regional restrictions and detection systems.
In a separate report, Anthropic said Russian- and Chinese-speaking operators used its models to automate cyberattacks. A Russian-speaking operator named "JackPoterz" utilized customized AI-driven workflows targeting more than 20 organizations, including government ministries, embassies, and diplomatic missions in Ukraine and Europe. One workflow produced more than a dozen possible zero-day findings in one month.
Additionally, a likely Mali-based consultant worked with the country's state intelligence service to build a population-scale domestic surveillance system monitoring roughly 25 million SIM cards. This platform utilized on-premises models with Claude providing software design and engineering support.
