OpenAI Catches Chinese AI Firm Stealing Model Data

1 min read

OpenAI has uncovered a large-scale effort by associates of a Chinese AI firm to extract the ‘hidden reasoning’ behind its AI models, according to a company report. The lab identified more than 15,000 users attempting to copy its work, with a core group linked to Moonshot AI, the creator of the Kimi chatbot.

The distillation activity, which began on July 1, reached its peak on July 24 and 25, when over 4,000 users sent approximately 16,000 requests. OpenAI took action to shut down the operation by July 28, limiting access to fake accounts and making it harder to create new ones. The company emphasized that the technique could allow rivals to build copycat models without the safety protections OpenAI includes, posing a risk to national security.

Thomas Duesterberg, a senior fellow at the Hudson Institute, noted that the findings highlight how Chinese AI companies rely on American technology to remain competitive. He stated that the Chinese firms ‘can’t compete at all unless they steal our stuff.’ According to Duesterberg, the operators targeted the core of OpenAI’s models rather than ordinary data, which he described as a ‘polite way of saying they’re stealing the stuff.’

Duesterberg also warned that Chinese AI models are designed to repeat government-approved narratives, potentially misleading American users on political matters. He urged Washington to take stronger measures, suggesting a ban on the use of these models in the U.S. and a restriction on U.S. capital investments in Chinese firms.

OpenAI is not the only company facing such accusations. In January 2025, DeepSeek, another Chinese AI firm, was accused of potentially using OpenAI’s models to train its own, according to Axios. At the time, former White House AI adviser David Sacks stated there was ‘substantial evidence’ of distillation.

Source: Daily Caller

Written by

Ryan Wilson