Illustration of WhatsApp’s limited-beta Scam Alert warning shown inside a suspicious non-contact chat.
WhatsApp is testing an optional Scam Alert feature that flags suspicious chats from non-contacts on the user’s device, while keeping message content on the phone and preserving end-to-end encryption. The feature is still in limited beta rather than broad release, but its purpose is straightforward: warn people about likely scam and social-engineering attempts before they reply.
Scam Alert flags suspicious non-contact chats on the phone
Scam Alert is aimed at a problem WhatsApp users already know well: increasingly convincing scam outreach that often starts with an unexpected message from someone outside a saved contact list. Instead of sending those messages to Meta for review, WhatsApp says the classification happens locally on the device. That means the feature can analyze likely scam patterns in inbound non-contact chats without moving message content off the phone for detection.
For sales, support, or intake teams that handle first-contact messages on WhatsApp, that means a suspicious inbound chat may be flagged before anyone replies. That can help staff avoid engaging with fraudulent outreach while keeping customer screening on the device instead of routing message content to central review. The feature is also optional, not default-on, so users are not being pushed into a system they did not choose.
Inside the warning: block, report, continue, or trust
When WhatsApp flags a conversation, the warning appears inside the chat itself and only for the recipient. From there, the user gets four immediate options: block the sender, report the chat, continue anyway, or mark the conversation as trusted. That keeps the decision close to the message, instead of burying it in a separate moderation screen or account setting.
The trust option changes how the system behaves going forward. If a user decides a flagged conversation is legitimate and marks it as trusted, WhatsApp stops flagging that chat again. Meta also says the feedback it receives from the feature is limited to aggregate alert-frequency and accuracy data based on how users respond to warnings, not the content of the messages themselves.
The privacy and security controls around the rollout
WhatsApp is pairing Scam Alert with several technical controls meant to verify the model and protect how it is delivered. Meta says the system includes model-hash verification, an append-only transparency ledger, and OHTTP relays for model downloads. Meta has also tied model version verification to SHA-256 hashes, though it has not clearly shown where any related controls or details appear inside the app.
WhatsApp is also expanding its bug bounty coverage to include Scam Alert, including the model weights and the analytics pipeline. What remains unclear is the rollout scope: WhatsApp has not confirmed which regions, languages, or device types are included, and there is still no public launch date for wider availability.
For now, Scam Alert remains a limited opt-in beta with no public accuracy or false-positive data and no confirmed timeline for a wider release. If it reaches broader availability, WhatsApp users would get an on-device warning for suspicious non-contact chats before replying, with the same block, report, continue, and trust options inside the conversation.
Disclaimer: This article was created with the assistance of AI. Images are for illustrative purposes only.
About the author
Samarth Agrawal is an AI and technology professional who writes about WhatsApp, automation, and emerging AI trends. He focuses on simplifying complex tech updates into practical insights for businesses, creators, and everyday users
