Meta Tests AI Scam Alert Feature for WhatsApp

Meta is testing a new optional security feature for WhatsApp called Scam Alert, which uses on-device artificial intelligence to identify and flag fraudulent messages. The tool employs a machine learning model downloaded directly to the user's phone to analyze incoming messages from unknown senders for patterns typical of scams, including artificial urgency, requests for personal information, and romance-baiting tactics.

To preserve end-to-end encryption, the AI processes data locally on the device rather than on company servers. Message content is only sent to Meta if a user manually opts to share the last five messages of a trusted chat to improve model accuracy. When the system detects a potential scam, it displays a warning banner visible only to the recipient, who can then block the sender, report the conversation, or mark the contact as trusted.

The feature is currently in a limited beta phase for security researchers and early users to gather feedback and conduct stress tests for vulnerabilities. Users must manually enable the tool via settings. The initiative follows Federal Trade Commission reports stating that social media scams caused 2.1 billion dollars in losses in 2025, with 425 million dollars of those losses occurring specifically on WhatsApp.