Mar 15, 01:38 AM

Researchers Identify Side Channel Exposing AI Assistants, Google Gemini Exception with Encryption Measures by Various Companies

Researchers have identified a side channel that can expose encrypted responses from AI assistants, with Google Gemini being the exception. The real-time token transmission in chat-based LLMs plays a crucial role in this vulnerability. Hackers can exploit this flaw to read private AI-assistant chats, despite encryption measures implemented by various companies.

#Google Gemini

Written with ChatGPT (GPT-3).