← Azimuth Daily
Capturing data…
Capturing data…
FRI, 2 OCT · 93 ITEMS
Bloomberg, Financial Post, CNBC + 1 more · AI & Technology
4 outletsOpenAI publicly accused Chinese AI lab Moonshot of running a month-long campaign, primarily in July 2026, to extract proprietary reasoning from OpenAI's GPT models using a technique called — sending massive numbers of queries to siphon off enough outputs to replicate the model's behavior. This is part of a broader pattern: in February 2026, Anthropic had already flagged Moonshot and other Chinese labs for similar extraction attacks, and in September 2026, CISA issued an advisory naming Moonshot among several Chinese AI companies conducting industrial-scale extraction of AI model data.
The claim is solid, based on a direct announcement or briefing from OpenAI reported by Bloomberg, and promptly corroborated by multiple other reputable outlets (CNBC, Financial Post, Semafor). External evidence from Anthropic's earlier detection (February 2026) and CISA's advisory (September 2026) independently confirm that Moonshot was involved in systematic distillation campaigns, so the accusation fits a well-documented pattern of activity.
Genuinely new reporting from September 30 – October 1, 2026; no indication that the underlying material is older or resurfacing.