Artificial Intelligence
OpenAI accuses Moonshot AI of reasoning-extraction campaign
OpenAI said it found a coordinated effort linked to Moonshot AI that copied model reasoning, blocked the accounts involved, and shared the findings with other labs and government channels.
- OpenAI said people linked to Moonshot AI were behind a coordinated effort to extract the reasoning of its models.
- The company said the activity began on 1 July, spiked on 24 and 25 July, and was shut down by 28 July.
- OpenAI said the operators copied encrypted reasoning from one conversation and used another model to decrypt and transcribe it.
- Caroline Zier said OpenAI’s concern was about breaches of its terms, and the company said it had shared its findings with the Frontier Model Forum and government channels.
OpenAI said in a blog post on Wednesday that people linked to Moonshot AI had carried out a coordinated effort against its models. It said it could not connect every operator to one actor, but attributed a core group to people associated with Moonshot.
The company said the activity began on 1 July at low volume and rose sharply on 24 and 25 July, when it saw 16,000 requests from more than 4,000 users. OpenAI said related activity later appeared across more than 15,000 users before it stopped the campaign on 28 July.
OpenAI said the operators copied encrypted reasoning from one conversation and then used another model in a separate conversation to decode and transcribe it. It said they did not break the encryption, enter a database, or access stored user conversations.
The company said it had banned accounts, tightened sign-up checks, and added protections for hidden reasoning. Caroline Zier, who leads strategic national security policy initiatives at OpenAI, told Bloomberg that the concern was a terms-of-service breach rather than open models or legitimate distillation.
Bloomberg reported that this was the first time OpenAI had accused Moonshot AI. It also said Anthropic had made a similar claim last month, US agencies had issued the same charge, China had rejected that advisory, and David Sacks had criticised the reports as a push to ban rival open models.
Named in this story
People
- Caroline Zier
- said OpenAI’s concern was about terms-of-service breaches
- David Sacks
- criticised such reports as a push for a ban on rival open models
Companies
- OpenAI
- said people linked to Moonshot AI ran the campaign and then blocked accounts
- Moonshot AI
- was accused of being behind the extraction effort
- Anthropic
- had made a similar claim against Moonshot AI
- DeepSeek
- was named alongside Moonshot AI in Chinese regulatory scrutiny
Organisations
- Frontier Model Forum
- received OpenAI’s findings
Governments and agencies
- US agencies
- had made the same charge in an advisory
- China
- rejected the advisory as unfounded accusations
- Chinese regulators
- were reported to be investigating DeepSeek and Moonshot AI
Products and systems
- Kimi
- was named as Moonshot AI’s model
How the source tells it
The piece was mostly straight reporting, but it used OpenAI’s allegation of coordinated extraction and the article’s own technical terms to give the story a mildly accusatory edge.
- accusation the account rested on OpenAI’s claim of a coordinated campaign and repeated its technical description of the conduct
- pushback it balanced the allegation with rival claims, scepticism, and stated denials from other parties