OpenAI Identifies Moonshot AI Campaign to Extract GPT Reasoning Capabilities

OpenAI has publicly accused Moonshot AI, maker of the Chinese chatbot Kimi, of orchestrating a large-scale campaign to extract the hidden reasoning and capabilities of its artificial intelligence models through a technique known as adversarial distillation. The U.S. AI lab said it discovered and disrupted the coordinated effort in late July, which generated 16,000 requests from over 4,000 users across a 48-hour window and involved related activity across more than 15,000 accounts total.