Technology
OpenAI says it disrupted Moonshot-linked campaign to extract model reasoning
The company attributes a core cluster of “adversarial distillation” traffic to people associated with China’s Moonshot AI, developer of Kimi, after spikes of thousands of extraction-pattern requests in July.
Published: October 1, 2026 · 1 min read
OpenAI said it identified and disrupted a coordinated campaign to extract protected reasoning from its models — activity it described as adversarial distillation — and attributed a core cluster of that traffic to individuals associated with Moonshot AI, the Chinese startup behind the Kimi models.
In a Sept. 30 blog post, OpenAI said the earliest observed activity began in the first week of July and that high-volume spikes on July 24–25 included 16,000 requests using a relevant extraction pattern from more than 4,000 users. Related prompt-pattern activity spanned a cluster of more than 15,000 users before the company said it fully disrupted the campaign by July 28. Operators did not break encryption or access databases or stored conversations, OpenAI said; instead they manipulated model interactions so hidden reasoning could be reproduced in a form visible to the requester, including by copying encrypted reasoning between conversations and asking a model to transcribe it.
OpenAI said it was unclear whether all operators came from a single actor, but that a core cluster was linked to people associated with Moonshot. The company said it has shared findings through the Frontier Model Forum and government channels, closed a pathway that allowed replay of another user’s encrypted reasoning, and tightened account and infrastructure controls. Moonshot has previously faced distillation allegations from Anthropic; OpenAI’s post is its first public attribution tying Moonshot-linked accounts to this July campaign.
The episode sharpens the commercial and national-security fight over whether rivals can shortcut frontier training by harvesting another lab’s hidden chain-of-thought — and puts pressure on both U.S. and Chinese labs to prove that product gains are not built on unauthorized extraction.
Sources: OpenAI, “Disrupting a coordinated model-distillation campaign”; CNBC; Financial Post / Bloomberg.
Sources
Newsletter
News. Context. What matters.
One essential briefing, written for people who would rather understand the story than scroll it.
Unsubscribe anytime. We don’t sell addresses.
Recommended
Technology
Signal, Meta and Tailscale face senators tonight over Ottawa's bill to force apps to build intercept tools
The Senate national security committee hears from the tech companies at 6 p.m. ET Monday as it weighs Bill C-22, which would let the public safety minister order providers to build lawful-access capabilities and keep user metadata for up to six months.
Technology
Italy fines Meta $887 million for reading Android users' photos without consent
Italy's privacy regulator said Facebook accessed the photo galleries of about 20 million Android users without specific consent. Meta says it will appeal.
Technology
Microsoft 365 went down worldwide — Outlook, Teams and Copilot hit for more than 18 hours
A cascading failure that began Thursday evening left email and workplace tools unreachable across more than 70 countries. Microsoft says it was not a cyberattack — and service was still recovering on Friday.