WhatsApp has announced a new security feature aimed at helping users verify the identity of the people they communicate with, especially if they are not registered in the contacts list.
The new tool, called "Scam Alert," relies on artificial intelligence to analyze messages received from unknown numbers, studying "conversational structure and linguistic cues" to assess the likelihood that the message is part of a scam
When the system suspects fraudulent intent, a warning appears within the conversation that is hidden from the other party, and allows the user the freedom to choose to block the sender, report him, or continue the conversation as usual.
The user needs to manually activate the feature through the settings. Once activated, the application downloads a machine learning model that runs locally on the user's device, meaning that Meta, the parent company of WhatsApp, does not receive any copy of the chat content, nor does it share it with any third party, according to the company's assurances.
However, WhatsApp confirms that it collects certain technical data in a completely anonymous manner, in order to ensure the feature's effectiveness and to keep it updated to keep pace with the latest fraud methods.
The fraud alert system was built based on previous conversations reported by users as fraud attempts, and these were used to train an AI model to recognize suspicious messages.
The feature is currently in a limited beta version for a group of early users, and WhatsApp plans to conduct extensive testing of the system and has invited researchers to report any vulnerabilities, before it is rolled out to all user at a later stage.
