OpenAI Accuses Chinese AI Firm of Copying Its Technology

An AI lab from the U.S. claims that associates of a competing Chinese company attempted to replicate its proprietary methods.

OpenAI reported on Wednesday that it had to intervene in a situation where over 15,000 users were trying to uncover the “hidden reasoning” behind its AI systems. The company links this effort to a group closely associated with Moonshot AI, the creator of the Kimi chatbot. They cautioned that this method could allow competitors to create imitation models lacking the safety measures integrated into OpenAI’s own systems, which poses a threat to national security.

“It’s not entirely clear if all the operators we observed during this period were from the same source. Still, we connect a significant portion of the activity to individuals tied to Moonshot AI,” OpenAI mentioned in its findings.

The AI creator took steps to shut down or restrict the accounts involved, made the signup process for new accounts more difficult, and is now actively monitoring for similar attempts, as indicated in the report.

The distillation activities began on July 1 and reached their peak on July 24 and 25, with around 4,000 users sending out approximately 16,000 requests, according to the findings. OpenAI managed to put an end to it by July 28.

Neither Moonshot AI, OpenAI, nor the Chinese Embassy responded promptly to requests for comments.

Thomas Duesterberg, a senior fellow at the Hudson Institute specializing in U.S.-China economic relations, suggested that these findings reveal a reliance among Chinese AI companies on American technology to remain competitive.

“This shows that Chinese firms, which claim to rival American corporations, really can’t compete unless they resort to stealing our technology,” Duesterberg explained.

He noted that the operatives targeted OpenAI’s core models rather than just regular data, referring to the infringement of OpenAI’s terms as “a polite way of saying they’re stealing.”

Furthermore, Duesterberg warned that Chinese AI models are programmed to convey government-approved narratives and might not address discussions about significant events like Tiananmen Square. This could lead American users to be misled on political issues.

“They are systematic and relentless,” he added.

Duesterberg believes that the U.S. should take more decisive actions beyond just banning accounts. “I think we ought to prohibit the use of those models in the United States and restrict American investments in Chinese firms,” he stated.

Another Chinese AI company, DeepSeek, faced similar allegations in January 2025 when OpenAI examined whether it had used its models for training. At that time, former White House AI advisor David Sacks asserted that there was “substantial evidence” of such distillation.