WhatsApp Trials Scam Alert Feature to Detect Fraudulent Messages

WhatsApp has announced a new security feature called Scam Alert, designed to help users recognise scam or fraudulent messages. The feature uses a machine learning model that runs directly on the user’s device.

“This feature complements end-to-end encryption while recognising scams. It is optional and only works after the user activates it,” said WhatsApp’s parent company, Meta, in a statement reported by Forbes on Sunday (16/8/2026).

Once activated, the machine learning model is downloaded to the user’s device. The model then runs an inference process to classify whether incoming messages from non-contacts match known scam patterns.

Meta stated that the model is small and relatively simple enough to run directly on the device without requiring server-side components.

“No message content leaves the device for classification or is automatically reported to WhatsApp, Meta, or any other party,” said Meta’s technical team in its announcement.

The AI model in Scam Alert is trained using scam patterns previously reported by users to WhatsApp. Classification is carried out probabilistically, taking into account conversational structure and linguistic signals.

This means the system attempts to identify communication patterns resembling known scam methods, rather than relying solely on a single word or phrase.