Americans lost $425 million to scams on WhatsApp in 2025 alone. Meta is now testing a tool it says can help stop those losses before they happen.
The company announced this week that it is rolling out a new optional feature called Scam Alert for WhatsApp users. The tool uses an on-device machine learning model to identify messages that may be scams. According to CNET, the feature is currently in limited beta testing.
The tool works by running entirely on the user's phone. WhatsApp, Meta and any outside parties do not receive access to the contents of private conversations. Users must turn on Scam Alert through WhatsApp's settings menu, after which the AI model is automatically downloaded to the device. The feature can be turned off at any time.
Once active, Scam Alert runs in the background and scans incoming messages from unknown senders. It looks for patterns associated with scams, including requests for personal information, urgent language and deliberate trust-building tactics used in what are called romance-baiting scams, in which a scammer builds a fake emotional relationship with a target before requesting money.
When the tool flags a message as a potential scam, a warning banner appears that only the recipient can see. The user then has several options: block the sender, report the conversation, ignore the warning or mark the sender as trusted, which removes the flag and stops future alerts for that exchange.
Because the model is designed to improve over time, users can choose to share the last five messages from a trusted sender directly with WhatsApp to help keep the model accurate. That data sharing is opt-in only and does not happen automatically.
The Federal Trade Commission reported that consumers lost $2.1 billion to social media scams in 2025. WhatsApp is the most widely used messaging app in the world, with more than 3 billion people using it daily, making it a frequent target for scammers operating at scale.
Meta has not announced a timeline for the feature's full public release. A Meta spokesperson did not respond to a request for comment.
