Scammers have been targeting WhatsApp users for years, and the numbers prove it. The FTC says victims reported losing $425 million to scams on WhatsApp alone, part of a broader $2.1 billion lost to social media scams in 2025.
Now Meta is fighting back with Scam Alert, a new optional feature rolling out in limited beta that flags suspicious messages using artificial intelligence, all without ever reading your actual conversations.
How does Scam Alert catch scams without breaking encryption?
Scam Alert runs an AI model directly on your phone instead of on Meta’s servers. Once you turn it on, the app downloads that model and uses it to scan incoming messages from people outside your contacts, checking for patterns tied to known scams like conversational structure and specific language cues.
Meta says no message content ever leaves your device for this classification, and nothing gets automatically reported to WhatsApp or Meta unless you choose to report it yourself. This keeps the feature compatible with WhatsApp’s end-to-end encryption, since your private conversations will never reach Meta’s servers.
What happens when a scam gets flagged?
If the model catches a likely scam, you will see a warning inside that specific chat, visible only to you but not the sender. From there, you can block the sender, report the conversation, or simply continue chatting if you want.
If a warning turns out to be wrong, you can mark that chat as trusted, and Scam Alert will stop flagging it going forward. Whenever you mark a chat as trusted, you can also choose to share the last five messages with WhatsApp to help improve the model’s future accuracy.
The Scam Alert feature builds on Meta’s earlier scam detection work covering suspicious device linking requests, adding another layer of AI-powered protection across the platform.
