How We’re Building Scam Alert on WhatsApp With End-to-End Encryption and Verifiability Guarantees

Meta's WhatsApp Scam Alert design separates checking a message from reporting it. Once enabled, a downloaded model examines messages from non-contacts on the phone and shows a private warning when it detects a likely scam. The user decides whether to block, report or continue. Message text stays on the device for classification, according to the August 12 technical preview, which describes a limited beta.

Measuring the feature uses a separate path: devices contribute counts of warnings and user actions, which protected computing environments combine across users and add statistical noise to before releasing totals. Local classification therefore coexists with aggregate performance measurement.