Meta Tests AI-Based Scam Alert Feature on WhatsApp to Prevent Fraud

Meta continues to strengthen the security system on its instant messaging platform, WhatsApp. The technology giant is reportedly preparing a new feature called Scam Alert, designed specifically to combat the increasingly rampant modes of digital fraud or scamming.

The Scam Alert feature utilises artificial intelligence (AI) technology to help users recognise potentially harmful messages. The feature is currently in a limited beta testing phase before being released globally.

Unlike cloud-based systems, Scam Alert works by running a machine learning model directly on the user’s device (on-device). This is done to ensure privacy is maintained. According to a CNET report on Thursday (13/8), Meta guarantees that conversation data will not be shared with WhatsApp, Meta, or any third party.

To use this feature, users need to activate it manually through the settings menu. Once active, the system will download the AI model to the device and work in the background to analyse incoming text messages, especially from unknown senders.

The AI in this feature is trained to recognise suspicious communication patterns. Meta emphasises that control remains in the hands of the user. If the system incorrectly detects a message and the user chooses to trust that sender, the warning sign will be permanently removed for that conversation.

Furthermore, the process of sharing data for AI improvement does not occur automatically. Users must actively provide consent before data is sent to WhatsApp servers to help improve detection accuracy in the future.

To date, Meta has not provided a definite date for when Scam Alert will be available to all public users. The feature remains part of an intensive testing phase to ensure the system’s readiness in distinguishing between everyday communication and actual fraud attempts.