Ugh “Private AI Compute” isn’t private. It sends your conversations to Google’s servers for processing. They say it’s encrypted and not even they can read it, but how does the AI model know it’s a scam if the AI model can’t read it? Is it encrypted at rest but decrypted on Google’s servers every time you receive a message? That’s not private. You have to trust Google not to read it, and you can’t trust Google anymore.