Meta announced on Wednesday that it’s rolling out a new tool, Scam Alert, to thwart con artists on WhatsApp. The optional feature, currently in beta testing, will run an on-device machine learning model to flag scam messages without giving WhatsApp, Meta or any other entity access to your private conversations.
To take part in the feature, you must turn on the Scam Alert tool in WhatsApp’s settings menu. From there, the AI tool will be automatically downloaded to your phone. You can disable the feature at any time.
Scam Alert operates in the background by classifying incoming text messages from unknown senders. It looks at communication patterns associated with scams on the app, such as requests for personal information, an urgent tone and deliberate trust-building tactics used in romance-baiting scams, in which scammers create a fake emotional bond before asking for financial assistance.
WhatsApp will display a banner warning of a potential scam in a message that only the recipient will see. To move forward, you can either block the sender, report the conversation, ignore the warning or trust the unknown sender. That last option removes the warning flag and stops future alerts for that conversation.
Since this is an AI tool we’re talking about, it’s constantly learning, which means you can choose to share the last five messages from a trusted sender directly with WhatsApp to help keep the model accurate. This part of the feature doesn’t happen on its own; You must opt in to share this data.
Scam Alert is currently in limited beta, and Meta hasn’t yet announced the feature’s full public launch timeline.
A Meta spokesperson didn’t immediately reply to a request for comment.
WhatsApp is the most popular messaging app in the world, with over 3 billion people using it daily. The Federal Trade Commission reported that, in 2025 alone, consumers lost $2.1 billion to social media scams. On WhatsApp specifically that year, Americans lost $425 million.