NewsIncident
OpenAI disrupts a reasoning-extraction campaign linked to Moonshot AI associates
From July 1 to July 28, 2026, accounts tied to people associated with Beijing-based Moonshot AI replayed encrypted reasoning traces from OpenAI's advanced models into weaker models of the same provider, which decoded them in plaintext. Activity peaked on July 24 and 25 at about 16,000 requests from over 4,000 users. OpenAI banned the accounts and blocked replay of encrypted reasoning across sessions.
Why it matters
The attack used no break-in, only an interface that trusted data across sessions and models. Model APIs need the same threat modeling as any other API.