OpenAI disrupted a campaign in which operators manipulated model interactions to expose protected internal reasoning traces that are normally encrypted, a technique the company called adversarial distillation. Activity spiked in late July 2026, with roughly 16,000 requests from more than 4,000 accounts over two days, and a broader investigation identified related activity across more than 15,000 accounts. OpenAI attributed a core cluster of the activity to individuals associated with Moonshot AI, the Beijing-based developer of the Kimi chatbot, closed the underlying mechanism, and began blocking associated accounts.