Meta adds AI screening to detect WhatsApp scams
The Verge Stevie Bonifield
Meta is testing a Scam Alert for WhatsApp that flags sketchy chats on the phone itself. It can warn users before they get burned, and the other person won’t see the alert.
Based on reporting by The Verge, Stevie Bonifield — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Meta is adding a new optional Scam Alert feature to WhatsApp, using on-device machine learning to spot messages that look suspicious. The rollout is starting as a limited beta, so this is a test run rather than a full launch.
When the model thinks a message could be a scam, WhatsApp shows a warning inside the chat. The other person can’t see that alert, which keeps the response quiet and one-sided.
From there, the user gets a simple set of choices: block the chat, report it, or keep talking. If the warning looks wrong, the user can also say so.
This isn’t Meta’s first pass at scam defense on WhatsApp. Earlier this year, the company launched detection for device linking requests, another place where scammers try to slip through.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI feature that doesn’t sound like a demo in search of a problem. Scam filters on-device make sense because the first line of defense should be boring and local, not a cloud party with your chats. Meta still gets to look heroic for cleaning up a mess it helped scale, which is the most Meta sentence imaginable.
Read more about this at: The Verge