Is It All a Scam? Anthropic Believes Chinese Competitors Are Cheating

Source: dpa 2 min Reading Time

Related Vendors

The AI company Anthropic accuses its Chinese rival of secretly redirecting user queries to it instead of answering them with its own software. This could be a bombshell...

If what Anthropics is accusing its Chinese AI competitors of is true, then the industry is in for quite a shakeup! At stake is nothing less than Chinese AI systems—such as Kimi and Deepseek—that were actually trained indirectly using U.S. AI systems.(Image:  AI-generated)
If what Anthropics is accusing its Chinese AI competitors of is true, then the industry is in for quite a shakeup! At stake is nothing less than Chinese AI systems—such as Kimi and Deepseek—that were actually trained indirectly using U.S. AI systems.
(Image: AI-generated)

According to Anthropic, the rerouting of requests has, in some cases, resulted in data from government agencies in China and Russia ending up in the U.S. company’s systems. Anthropic has already published a detailed report on attempts at unauthorized use. Here’s an example: Someone uploaded footage from surveillance cameras (including footage from areas near Chinese military bases) with the aim of analyzing the behavior of a specific person in the videos. The user assumed that the request would be processed by the Chinese AI software Kimi, developed by Moonshot. However, Kimi rerouted the request and had it handled by Anthropic’s Claude, according to the report.

Abuse of Hundreds of Thousands of Requests

According to Anthropic, in just a ten-day period, Kimi had Claude respond to nearly 300,000 customer inquiries. A network of 5,380 fake accounts was used for this purpose. According to Anthropic, the even better-known Chinese AI Deepseek also drew attention with such tricks. Among other things, this led to inquiries being redirected to the U.S. company containing data from an agency affiliated with the Russian Ministry of Defense. This data also included valid login credentials for a Russian government database. The allegations could now cast a shadow on the reputation of the Chinese AI industry.

Did US-AI Actually Train Kimi & Co.?

Since the beginning of last year, low-cost AI models—first from Deepseek and then from Moonshot—had repeatedly caused turmoil on the stock market. And at times, investors doubted the business prospects of American AI labs such as Anthropic and “OpenAI,” the developer of “ChatGPT,” because it appeared that powerful artificial intelligence could be trained with significantly less effort. The U.S. side is now accusing the Chinese developers of having used American software on a large scale to train their AI. According to the report, Kimi alone accessed Anthropic’s AI 23 million times between May and July using this technique, known in the industry as “distillation.”

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta)

Unfold for details of your consent