ISLAMABAD: Meta announced a new feature for WhatsApp that automatically detects suspicious messages using artificial intelligence that runs on users' phones rather than on company servers.

According to Forbes, the Scam Alert feature began rolling out to a limited beta group through Meta's bug bounty research program. The company has not announced a timeline for the release of the feature for the users.

When enabled, the feature examines incoming messages from contacts not saved in a user's phone and checks for patterns associated with known scams. If the system detects a suspicious message, it displays a warning on the user's screen. The sender receives no notification.

The users can then block the contact, report the message, continue the conversation or mark the sender as trusted.

All analysis happens on the device itself. Meta said message content does not leave the phone for processing.

Scams on WhatsApp have increased significantly. Meta disabled over 6.8 million WhatsApp accounts linked to organized crime and scam centers in 2025, as per BBC. Over 90% of all phishing attacks executed via messaging apps are delivered through WhatsApp. The US Federal Trade Commission reported that fraud originating on WhatsApp cost Americans $425 million in 2025 alone.

The Global Anti-Scam Alliance, an international organization dedicated to cross-border fraud analysis, reported that 13% of global adults encounter an active digital scam attempt every single day, based on multi-market studies analyzing data across tens of thousands of victims.

online scams

Meta deploys on-device AI to detect scams on WhatsApp