OpenAI Accuses Moonshot AI of Extracting Data for Model Training
On September 30, OpenAI claimed that it identified a coordinated campaign aimed at extracting protected reasoning from its models and traced much of the activity to Moonshot Intelligence, the Chinese company behind Kimi. While not directly linking all actions to one entity, OpenAI highlighted a core cluster connected to Moonshot. Anthropic has also raised similar concerns about Moonshot's activities. The allegations involve what is known as model distillation — an approach where data from another AI system is used to train new models without direct access to the original training set. This could potentially undermine proprietary protections and intellectual property rights in AI development, raising broader questions around ethical practices and security measures within the industry. Such accusations may lead to increased scrutiny of Moonshot's operations and impact its reputation as well as investor confidence in companies involved with similar data extraction techniques.
Topics: data extraction model distillation Moonshot AI OpenAI