OpenAI Disrupts AI Model Extraction Campaign Linked to Moonshot AI
OpenAI reported identifying and disrupting an organized campaign where operators attempted to extract protected reasoning from its artificial intelligence models, linking a core cluster of the activity to Chinese startup Moonshot AI. The effort involved over 16,000 requests across more than 4,000 users in two days, with related activity spanning over 15,000 users before being fully disrupted by July 28. The campaign began in early July and is part of a broader concern among U.S. AI companies about rivals using their models to accelerate development without equivalent investment or safeguards. OpenAI described the method as 'adversarial distillation,' where outputs from one model are used to train another, potentially compromising safety and national security. OpenAI emphasized that operators did not breach its encryption, databases, or user conversations but manipulated interactions with AI models instead. The company shared findings through industry forums and government channels.
Tickers: $BABA
Topics: adversarial distillation AI model security cyber threats Moonshot AI OpenAI