Google Reports Widespread Attempts to Clone its Gemini AI

Google's Threat Intelligence Group blocked a coordinated campaign to copy Gemini's proprietary reasoning capabilities. The company identified over 100,000 prompts linked to model extraction or distillation attempts. These attempts aim to create smaller models that mimic larger ones. Google states these actions violate terms of service and may trigger legal action.

Google identified a new malware family named HONESTCUE. This malware integrates the Gemini API to generate malicious code.

The company shut down a phishing kit named COINBAIT. AI code generation platforms powered the COINBAIT kit. Google also disabled a service falsely advertising custom malware-generating AI.