New WhatsApp Security Feature Uses AI to Detect Scam Messages

Meta has begun testing a new security feature on WhatsApp called “Scam Alert,” designed to help users identify suspicious messages and reduce scam attempts. The feature is currently being tested on a limited basis among users of the Android beta version.

The feature analyzes incoming messages from unknown contacts and detects patterns commonly associated with scams. When suspicious indicators are identified, a warning appears in the conversation informing the recipient that the message “may be a scam.”

WhatsApp says the warning is visible only to the recipient, without notifying the sender. Users are then given several options for dealing with the suspicious conversation, including blocking or reporting the contact or marking the conversation as trusted.

The feature relies on a machine-learning model that runs directly on the user’s device to analyze messages. This means that conversation content is not sent to WhatsApp or Meta servers for analysis. Users may, however, optionally share certain data to help improve the system’s performance and reduce false classifications.

The feature also includes a local log allowing users to review when warnings were triggered, which messages were classified as suspicious, and the version of the AI model used for the analysis. The feature is disabled by default and must be activated manually.

The “Scam Alert” test comes amid a rise in scam attempts through messages and calls, particularly as fraudsters increasingly use artificial intelligence tools to create more convincing messages.

WhatsApp also plans to expand the feature to iPhones, although no official date has been announced for its wider rollout.