WhatsApp is now testing a new AI backed feature that will help users identify potential scam messages. It will be called Scam Alert and debut as an optional feature for the users. Currently, the feature is available for the limited number of Android beta users. For the unversed, the feature uses an on-device machine learning model to detect patterns commonly associated with scams. Importantly, the feature does not require WhatsApp to scan or send users’ messages to Meta’s cloud servers
As per the details shared by Meta, Scam Alert works by downloading a machine learning model to the user’s smartphone after the feature is enabled. The model then analyses incoming messages from unknown contacts and checks whether they match patterns associated with known scam attempts.
If a message is identified as potentially suspicious, WhatsApp displays a warning inside the chat. The warning is only visible to the recipient, meaning the sender will not know that the conversation has been flagged.
The users can then choose to block or report the sender or continue the conversation. And in case, the warning turns out to be incorrect, users can mark the chat as trusted. The Meta-owned platform will then stop flagging that particular conversation.
If you have already heard about the feature, the first thing that may have come to your mind is privacy. Meta says that the machine learning model and the messages it analyses remain on the user’s device. WhatsApp does not automatically send flagged messages to its servers, nor does the feature automatically report a user.
The users can voluntarily report a suspicious conversation to WhatsApp. They can also choose to share their last five messages from a chat to help improve the accuracy of the feature. However, this requires action from the user.
It should be noted that the feature is currently in testing and will not be available to the general public for several weeks. Once it is rolled out, it will reportedly be accessible via Settings > Account.