by Dylan Kresak
An American frontier artificial intelligence lab says it caught associates of a rival Chinese firm trying to copy its homework.
OpenAI said Wednesday it shut down an effort by more than 15,000 users to extract the “hidden reasoning” behind its AI models, tying a core group of people associated with Chinese-based Moonshot AI, the creator of the Kimi chatbot, according to a company report. The company warned that the technique could let rivals build copycat models without the safety protections OpenAI incorporates into its own, which risks national security.
“It is unclear whether all operators we observed during the relevant time period originated from a single actor. However, we attribute a core cluster of the activity to individuals associated with Moonshot AI, the developer of Kimi,” OpenAI said in its report.
The ChatGPT creator said it shut down or limited the fake accounts, made it harder to sign up for new ones and is watching more closely for similar activity, the report stated.
The distillation activity began July 1 and peaked July 24 and 25, when more than 4,000 users sent roughly 16,000 requests, according to the report. OpenAI shut the activity down by July 28.
Moonshot AI, OpenAI and the Chinese Embassy each did not immediately respond to the DCNF’s request fro comment.
Thomas Duesterberg, a senior fellow at the Hudson Institute who studies U.S.-China economic competition, said the findings show Chinese AI companies are leaning on American technology to keep up.
“[T]his indicates that the Chinese firms, which claim to be almost as powerful as the American firms, really can’t compete at all unless they steal our stuff,” Duesterberg told the Daily Caller News Foundation.
He said the operators went after the core of OpenAI’s models rather than ordinary data, calling the violation of OpenAI’s terms of use “a polite way of saying they’re stealing the stuff.”
Duesterberg also warned that Chinese AI Models are built to repeat government-approved narratives and will not answer questions about topics such as Tiananmen Square, meaning American users could be “led astray, or totally misled” on political matters.
“[T]hey’re disciplined and they’re relentless,” he said.
Duesterberg said Washington should go further than OpenAI’s account bans.
“I think we ought to ban the use of those models in the United States, and I think we ought to ban U.S. capital investments in Chinese firms,” he told the DCNF.
DeepSeek, another Chinese AI firm, faced a similar accusation in January 2025, when OpenAI said it was looking into whether the company may have used its models to train its own, Axios reported. Then-White House AI adviser David Sacks said in a Fox News interview at the time that there was “substantial evidence” of distillation.
The findings come as the White House was gathering the same firms to set the terms of the race. On Tuesday, President Donald Trump hosted executives from OpenAI, Anthropic, Nvidia, Meta, Google, Amazon and SpaceX for what he called an “extraordinary gathering” on “super intelligence.”
OpenAI president Greg Brockman was among those who signed a voluntary “White House Accord on Super Intelligence” committing the companies to internal controls, independent audits and board-level review of whether models stay within intended bounds. Trump told reporters the companies were “policing each other” and that the accord was “morally binding,” while ruling out new statutory guardrails and saying the United States would not “put on brakes.”
Days earlier, Chinese President Xi Jinping had been in Washington for a state visit in which artificial intelligence was on the agenda but model copying was not resolved. Trump and Xi agreed to open a bilateral channel for AI-related incidents and to hold a government dialogue on risks and benefits by November.
Xi said both countries had a responsibility to keep the technology “always under human control.”
Trump said afterward that he did not want to integrate U.S. and Chinese efforts because American firms were ahead by “at least a year,” and that Washington was still pressing Beijing over Chinese companies copying American models. Chinese Ambassador Xie Feng later told Newsweek the two sides could “compete” but should “jointly prevent the misuse and abuse” of the technology — a formulation that does not address the distillation OpenAI says it shut down.
– – –
Dylan Kresak is a reporter at Daily Caller News Foundation. Executive Editor of The Tennessee Star and The Star News Network Christina Botteri contributed to this report.
