Anthropic has accused Chinese artificial intelligence company Moonshot AI of secretly routing user requests through Claude, its flagship AI model, without proper authorization. The allegations, reported by Bloomberg, suggest Moonshot may have been presenting Claude's outputs to its own users while obscuring the underlying model powering the responses. Anthropic confirmed the situation directly to Bloomberg, though the full scope of the arrangement remains unclear.
What Anthropic Is Alleging
According to the Bloomberg report, Anthropic says Moonshot AI accessed Claude through its API and funneled user queries through the model in a way that was not disclosed to end users. This would mean Moonshot customers believed they were interacting with Moonshot's own technology when they were, at least in part, receiving responses generated by Claude. Anthropic has not publicly detailed when the behavior was discovered or how long it may have continued before the company took action.
Key Facts
- Anthropic says Moonshot AI secretly routed user requests through the Claude API.
- Users were allegedly unaware their queries were being processed by Claude.
- Anthropic confirmed the allegations to Bloomberg but has not specified the timeframe.
- Moonshot AI is a prominent Chinese AI startup backed by major investors.
- The incident raises questions about API terms of service enforcement across the industry.
The case highlights a growing challenge for AI model providers: ensuring that companies accessing their APIs comply with usage policies, particularly around transparency and disclosure. Anthropic's terms of service prohibit customers from misrepresenting the origin of AI-generated content. Whether Moonshot's conduct constitutes a clear violation, and what remedies Anthropic is seeking, have not been disclosed. This story also arrives against a broader backdrop of scrutiny around how AI firms handle user data and model access, an issue that has drawn attention well beyond this particular incident. Those following earlier controversies involving Anthropic and user trust will find the themes familiar, even if the parties involved are different this time.
Anthropic confirmed to Bloomberg that Moonshot AI had been routing user requests through Claude without proper authorization, in apparent violation of its API terms of service.Bloomberg
Moonshot AI and Its Market Position
Moonshot AI, the Beijing-based startup behind the Kimi AI assistant, has positioned itself as one of China's leading homegrown AI companies. It has attracted significant investment and built a sizable user base across consumer and enterprise applications. If the Bloomberg allegations are accurate, the company was supplementing or replacing its own model outputs with those from a U.S.-based provider while keeping that fact from users. That would be a significant reputational issue at a time when Moonshot is competing hard for market share inside China.
For Anthropic, the incident underscores the difficulties of policing a sprawling API ecosystem. Claude's model family is now available to a wide range of developers and enterprise customers globally, and monitoring downstream usage for misrepresentation at scale is not straightforward. The company has not said whether it has terminated Moonshot's API access or taken other formal steps in response.
This episode will likely add pressure on AI companies to strengthen API auditing and to require clearer disclosure from business customers about how model outputs are presented to end users. It also raises questions about competitive dynamics: if companies can quietly build products on top of a rival's model and present the results as their own, the incentives for investing in original model development erode. Whether regulators in the U.S., China, or elsewhere will take an interest remains to be seen, but the story has already put a spotlight on a corner of the industry that tends to operate without much public scrutiny.
