OpenAI Disrupts Mass Reasoning Extraction Campaign Linked to Moonshot AI

OpenAI disclosed September 30 that it had detected and disrupted a coordinated campaign by individuals associated with Chinese AI startup Moonshot AI to extract protected model reasoning through adversarial attacks. The campaign, which peaked on July 24–25 with 16,000 extraction-pattern requests from more than 4,000 users across two days, represents one of the first large-scale attempts to systematically reverse-engineer an AI model's hidden reasoning capabilities.