Why Is Moonshot Using Claude Instead of Its Own Model?
A Chinese AI startup, Moonshot, has quietly swapped its domestic Kimi model for Anthropic’s Claude in its backend systems while continuing to gather user interactions to improve its own models. The shift, observed in September 2026, raises questions about the true origins of its technology and data practices.
Breaking news
Trump Orders Federal Agencies to Adopt „Super Intelligence” Terminology
The Silicon Valley Accountability Gap
Running a Local AI Model on a Phone Handles Most Chat Prompts
Google Photos Could Soon Offer a Fresh Start with Ask Photos FeatureThe change was noted by tech analyst David Agranovich, who highlighted the irony of a Chinese firm relying on an American AI system while presenting itself as a frontier innovator. Moonshot, known for its Kimi chatbot, appears to be routing user prompts through Claude, effectively using it as a proxy. Despite this, the company continues to log and store user exchanges, which are likely being used to train or refine its internal AI systems. This practice blurs the line between model deployment and data harvesting, especially given the geopolitical sensitivity around AI development and data sovereignty.
What Does This Mean for User Privacy and AI Independence?
Industry observers suggest that Moonshot may be using Claude as a temporary bridge while scaling its own capabilities, or that Kimi’s performance did not meet internal benchmarks. By integrating Claude, the company can maintain service quality while collecting real-world usage data. This data, including prompts and responses, could be instrumental in training future iterations of its proprietary models. However, the lack of transparency about this switch has sparked concern among users and regulators about informed consent and data usage.
The reliance on a foreign AI model, coupled with extensive data collection, complicates Moonshot’s positioning as a homegrown AI leader. Users may not be aware that their interactions are processed by Claude and potentially retained for training purposes. This situation underscores broader challenges in the global AI race, where companies navigate technological dependencies, data governance, and national security implications. Moving forward, Moonshot may face pressure to clarify its model architecture and data policies, especially as scrutiny over AI transparency intensifies worldwide.
Is Moonshot still using its Kimi model for any services? Current evidence suggests that Kimi has been replaced by Claude in Moonshot’s primary backend, though the company has not officially confirmed the extent of the switch.
Frequently Asked Questions
Are users informed that their data is being used for model training? There is no public indication that Moonshot has disclosed this practice to users, raising concerns about transparency and consent.
Could this affect Moonshot’s regulatory standing in China? Given China’s emphasis on AI self-reliance and data security, using a foreign model while harvesting user data could attract regulatory attention if deemed non-compliant with local AI governance rules.
