OpenAI Accuses Moonshot-Linked Users of Reasoning Theft

OpenAI says users linked to Moonshot AI ran a coordinated campaign to extract hidden model reasoning, starting 1 July and spiking 24-25 July with 16,000 requests from over 4,000 users. OpenAI shut it down by 28 July, banned accounts and tightened sign-up checks. It calls this adversarial distillation and has shared findings with other labs and government channels.