Google has reported that hackers are attempting to clone its Gemini AI model through "distillation attacks," a form of model extraction where adversaries systematically probe a mature machine learning model to gather information for training a new one. These attacks involve inundating Gemini with a massive volume of AI prompts, with one instance using over 100,000 prompts to steal the company's technology. Google's Threat Tracker report, published on Thursday, indicates that these efforts are likely originating from adversaries in countries such as North Korea, Russia, and China, and are part of a wider trend of AI-based attacks and malware. The company emphasizes that this threat targets service providers and model builders, not end-users, by risking the theft and replication of AI intellectual property.