Meta Introduces AI-Powered Scam Detection for WhatsApp
By Priya Anand

Meta has begun rolling out a new Scam Alert feature for WhatsApp, designed to identify and flag potentially fraudulent messages using on-device machine learning. This optional feature, currently available in a limited beta, alerts users to suspicious messages by displaying a private warning in the chat. The warning is visible only to the recipient, who can then choose to block, report, or continue the conversation.
Users have the option to mark a chat as trusted, thereby disabling future alerts for that conversation. If a chat is marked trusted, users can opt to share the last five received messages with WhatsApp to enhance the accuracy of the feature. Importantly, no message content leaves the user's device unless explicitly shared by the user, and the Scam Alert feature can be disabled at any time.
Earlier this year, Meta introduced additional anti-scam protections across its platforms, including warnings for suspicious device-linking requests on WhatsApp. These alerts are designed to prevent account hijacking by notifying users when a linking request appears fraudulent, such as requests for phone numbers, pairing codes, or fraudulent QR scans.
With over 3 billion users, WhatsApp is a significant target for scams, including sophisticated fraud schemes like wire transfers and "pig butchering" scams. In 2025, reported losses from WhatsApp scams amounted to $425 million, contributing to a total of over $2.1 billion lost to social media scams, underscoring the necessity for robust protective measures.