OpenAI accuses Moonshot AI of extracting encrypted reasoning via 16,000 requests
OpenAI has linked Moonshot AI, the lab behind the Kimi K3 model, to a coordinated campaign that attempted to extract encrypted reasoning from its models. In a post titled Disrupting a coordinated model-distillation campaign, OpenAI detailed efforts to bypass protections by manipulating model interactions across over 4,000 users, peaking at 16,000 requests on July 24–25. The campaign did not…
Key points
- OpenAI linked Moonshot AI to a 16,000-request campaign extracting encrypted reasoning via 4,000+ users
- Campaign began July 1, peaked July 24–25, disrupted by July 28 without database breaches
- Study shows Kimi K3’s responses mimic Anthropic’s Claude Opus 4.8, raising distillation concerns
The activity began July 1, with spikes tied to Kimi K3’s debut. OpenAI attributed the core cluster to Moonshot AI, echoing claims by the Trump administration and Anthropic earlier this year. Separately, a study found Kimi K3’s responses align unusually closely with Anthropic’s Claude Opus 4.8, suggesting potential distillation. Meanwhile, OpenAI’s newer models and Anthropic’s Claude Mythos remain vulnerable to reasoning extraction, while China’s Z.ai (Zhipu) GLM-5.3 model showed partial capability in hijacking control flows.
Model pages: Kimi K3 → · GPT-6 Astra → · GPT-6.1 Sol →
Moonshot AI Of Kimi K3 Fame Tried To Crack OpenAI’s Encrypted Reasoning Through 16,000 Requests, Bolstering Trump Administration’s Distillation Claims
wccftech.com · 30 September 2026
Loading the full article…
This text was published by wccftech.com and written by Rohail Saleem. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- Trump’s America.gov chatbot contradicts his claims on elections, wealth and wind turbines · 1 src
- Anthropic blocks scientists evading AI guardrails on gain-of-function research · 1 src
- Trump and tech CEOs sign voluntary AI safety accord with external audits · 14 src
- FTC launches probe into Anthropic, OpenAI over AI model risks · 10 src
- Bill Gates warns AI misuse could cause a billion deaths · 2 src
Comments
via GitHub Discussions