(Sorry if this is in the wrong category, didn't know exactly where this should go)
Microsoft recently disclosed a side-channel attack on AI chat bots called Whisper Leak. It allows an attacker who can see encrypted network traffic of streaming AI chat bots (like Kagi assistant) to detect accurately the topics of each chat solely based on packet size and timings.
Since Kagi focuses heavily on privacy, is the assistant vulnerable to this exploit? If so, are there any mitigations deployed?
The mitigation Microsoft deployed was just to add a random string to each response in their API.
