Scam & phishing detection
An on-device model reads a message the way a suspicious human would: who is it claiming to be, what does it want, and does the link agree with either answer?
- Couriers asking for a small "release fee"
- Banks manufacturing urgency about your account
- Payment links on domains that don't match the sender
- Family-in-trouble messages from unknown numbers
Flagged messages move to Spam and their links are disabled until you overrule it.