WhatsApp is preparing to introduce a new security feature designed to shield users from online scams and fraudulent activities. Meta initially shared details about this upcoming Scam Alert feature earlier this month, and the tool has now entered the testing phase for select Android users. This development indicates that the capability could soon roll out to a broader audience in the near future.
The primary objective of the scam alert tool is to identify potential fraudulent messages originating from unknown contacts. A notable aspect of this implementation is that WhatsApp will utilize an on-device machine learning model to analyze texts directly on the user smartphone rather than relying entirely on external servers.
Beta users get early access to Scam Alert
According to reports detailing the rollout, the Scam Alert feature is currently appearing for select users on WhatsApp Android beta version 2.26.34.1. The tool scans incoming messages from unrecognized senders to determine whether the text exhibits any known patterns or indicators associated with scams.
If the system flags a message as suspicious, a warning notification will appear directly within the relevant chat window. Following this alert, the recipient is presented with options to block the sender, report the account, or choose to continue the conversation regardless.
This feature is not mandatory by default, allowing users to enable it manually according to their preferences. The option can be located by navigating to WhatsApp Settings and selecting the Account section, where the Scam Alert toggle is made available.
On-device machine learning powers the detection
According to Meta, the scam alert mechanism relies on a machine learning model downloaded straight onto the user device. This model functions locally on the hardware, scanning incoming communications from unfamiliar numbers to spot suspicious phrasing and scam tactics.
When a message resembles a potential scam, the alert is displayed exclusively to the recipient. The sender remains unaware that their message has triggered a security warning on the other end.
In instances where WhatsApp incorrectly flags a legitimate chat as a scam, the user has the ability to mark the conversation as trusted. Once designated as trusted, that specific chat will no longer be flagged under the Scam Alert system.
User privacy remains a core focus of the implementation, with Meta confirming that message contents are not automatically reported to WhatsApp, Meta, or any third party during the detection process. The model has been trained on patterns observed in fraudulent exchanges, and users can optionally choose to share their last five messages to help improve the system. As testing continues among Android beta participants, a successful trial could lead to a wider global release, offering an extra layer of defense against unsolicited and deceptive messages.
