OpenAI says it disrupted a coordinated effort in which individuals connected to Chinese startup Moonshot AI allegedly attempted to extract the hidden reasoning processes behind its artificial intelligence models.
What OpenAI Uncovered
According to the company, the operation was far from small, touching more than 15,000 user accounts across its platform. OpenAI said it identified and shut down the activity, which it characterized as an organized attempt to probe the internal workings of its systems.
At the heart of the campaign, OpenAI traced a core cluster of accounts to people it associates with Moonshot AI, the Chinese firm behind the Kimi chatbot. The company suggested this group was specifically interested in the "hidden reasoning" layer of its models — the step-by-step logic that advanced AI systems generate internally before producing a final answer.
The hidden reasoning inside modern AI models has become one of the industry's most closely guarded secrets.
Why Hidden Reasoning Matters
The internal reasoning chains used by frontier models represent a significant competitive advantage. These processes allow AI systems to work through complex problems before delivering responses, and companies like OpenAI typically keep them concealed from users to protect proprietary techniques and prevent misuse.
Extracting or reconstructing those reasoning traces could help rival developers understand how a leading model arrives at its conclusions, potentially accelerating their own efforts to build competing systems. That makes the alleged activity particularly sensitive in an increasingly contested global AI race.
- The campaign reportedly spanned more than 15,000 user accounts.
- A central cluster was linked to individuals tied to Moonshot AI.
- The apparent goal was to access the models' concealed reasoning steps.
A Growing Pattern of Scrutiny
The disclosure adds to mounting tensions between US and Chinese players in the AI sector, where accusations of model copying and data extraction have grown more frequent. OpenAI has previously flagged attempts by various actors to misuse or reverse-engineer its technology.
Moonshot, valued among China's rising AI startups, has drawn attention for its Kimi product. The allegations, if accurate, underscore how competition over cutting-edge AI capabilities is spilling into disputes over intellectual property and the methods used to train and operate the most capable models.