Google Gemini Faces Large-Scale Model Distillation Attack, With Over 100,000 Prompts Leaking Core Logic in a Single Instance
Google's AI chatbot Gemini faced a large-scale 'distillation attack,' where attackers used over 100,000 repeated queries to extract its internal mechanisms, aiming to clone or enhance their own AI systems. Google attributed the attack to commercial motives, raising industry-wide concerns over large model security.....