OpenAI said on September 30, 2026 that it identified and disrupted a coordinated campaign to extract protected reasoning from its models, with the earliest observed activity in the first week of July. OpenAI described the activity as consistent with adversarial distillation — the unauthorized use of one model's outputs or reasoning to help train or reproduce another model. According to OpenAI, protected reasoning is a model's internal record for working through a task, and extracting it can reveal information withheld from final answers. The company said operators did not break encryption, compromise a database, or access stored user conversations, but instead manipulated model interactions so protected reasoning could be reproduced in visible forms. One method involved copying encrypted reasoning from one conversation and asking a model in another to decrypt and transcribe it. OpenAI reported the activity began July 1 at low volume, with spikes on July 24 and 25 totaling 16,000 requests from over 4,000 users, and related activity across more than 15,000 users that was fully disrupted by July 28. OpenAI said it attributes a core cluster of activity to individuals associated with Moonshot AI, developer of Kimi, while noting it was unclear whether all operators came from a single actor. The company said it banned accounts, strengthened controls, and shared findings through the Frontier Model Forum. The figures describe attempted, not necessarily successful, extractions.
- Earliest activity observed first week of July; spikes July 24-25 totaled 16,000 requests from over 4,000 users
- Related activity spanned over 15,000 users and was fully disrupted by July 28
- OpenAI attributes a core cluster to individuals associated with Moonshot AI, developer of Kimi
- No encryption was broken and no stored conversations were accessed, per OpenAI; figures describe attempted extractions
What it means for you
This is a dispute between AI labs about one company allegedly copying another's model behavior — it has essentially no bearing on how you use these tools day to day. OpenAI is the only source for the attribution claim, so treat the 'Moonshot AI' accusation as one company's account of events. Your own ChatGPT conversations were not accessed, per OpenAI.
Who should care
People who build products on top of AI model APIs, run security teams at AI companies, or follow the competitive dynamics between frontier labs closely.
Skip this if
You are an individual or small business simply using ChatGPT, Kimi, or similar tools to get work done — this changes nothing about your account, data, or workflow.
Sources: OpenAI — read the original