Microsoft Reveals 'Whisper Leak' Vulnerability in AI Chatbots

Microsoft has disclosed a new side-channel attack named 'Whisper Leak' that can allow eavesdroppers to infer the topics of conversations with AI chatbots, even when using end-to-end encryption. This vulnerability can be exploited by ISPs, governments, or individuals on the same Wi-Fi network by analyzing patterns in network traffic. Major AI providers, including OpenAI and Microsoft, have implemented defenses against this attack.