Google Reports on AI Model Extraction Attacks

Google Threat Intelligence Group (GTIG) has published a report highlighting attacks involving the extraction/distillation of AI models, where malicious actors and researchers exploit legitimate access to APIs to systematically probe models like Gemini AI. These attacks aim to replicate the logic and reasoning of the targeted models. The report indicates that these techniques are being used at all stages of cyberattacks, without specifying a date or quantified impact. No specific vulnerability (CVE) or malicious tool is mentioned.