Hackers can snoop on your private chats with AI bots
Microsoft has flagged a new vulnerability called "Whisper Leak" in popular AI chatbots like ChatGPT and Google's Gemini.
Basically, hackers can figure out what you're chatting about just by analyzing encrypted network traffic—no need to actually see your messages.
Even with encryption, the patterns and timing of data give away the topics users discuss.
Whisper leak could have serious implications
This flaw is especially worrying in places with strict surveillance or censorship, where authorities or internet providers could track conversations about sensitive issues like protests or elections.
Microsoft says attackers were able to guess topics almost perfectly and even piece together parts of chats.
They're urging AI developers to find ways to hide these telltale patterns as more people rely on chatbots every day.