WhatsApp has started testing a new Scam Alert feature designed to help users identify potentially fraudulent messages from unknown contacts.
The new security tool is currently being rolled out to a limited number of Android beta testers, marking another step in Meta’s efforts to combat scams on its messaging platform.
The feature uses an on-device machine learning model to analyse messages from people who are not saved in a user’s contacts and determine whether their conversational patterns resemble known scam activity.
How WhatsApp Scam Alert Works
Scam Alert is an optional feature that users can enable on supported versions of WhatsApp.
Once activated, WhatsApp downloads a machine learning model directly to the user’s device. The model then examines incoming messages from unknown contacts for patterns associated with potential scams.
Meta says the analysis takes place on the device rather than sending message content to WhatsApp or its parent company for classification. This approach is designed to allow scam detection while maintaining WhatsApp’s end-to-end encryption.
If the system determines that a conversation is likely to involve a scam, a warning is displayed to the recipient inside the chat.
The warning is private and is not shown to the person who sent the message.
Users can then decide whether to block the account, report the conversation or continue chatting.
Users Can Override Scam Warnings
WhatsApp is also giving users control over how the feature handles individual conversations.
If a user believes a warning has been triggered incorrectly, they can mark the conversation as trusted. The warning will then be removed, and Scam Alert will not flag that particular chat again.
Users who mark a conversation as trusted can also choose to share their last five received messages with WhatsApp to help improve the accuracy of the scam-detection system. This sharing is optional.
WhatsApp Uses AI Without Automatically Reporting Messages
One of the key aspects of the new feature is that WhatsApp says message content is not automatically sent to Meta, WhatsApp or a third party as part of the scam-classification process.
The machine learning model uses conversational structure and linguistic signals associated with scam messages.
According to Meta, the model was trained using patterns observed in scam conversations that users had previously reported.
This allows the system to identify characteristics that may indicate fraudulent activity while keeping the analysis on the user’s device.
Scam Alert Is Still in Beta Testing
The feature is not yet available to all WhatsApp users.
Reports indicate that Scam Alert is being gradually made available to selected Android beta testers, with compatible beta versions including 2.26.34.1 and later builds. Availability remains limited as WhatsApp continues testing the system.
WhatsApp has not announced a specific date for a full public rollout.
The company says it will continue testing and improving the feature before making it available more broadly.
WhatsApp Expands Its Fight Against Scams
The new Scam Alert feature forms part of Meta’s wider efforts to protect users from increasingly sophisticated online scams.
Earlier this year, Meta introduced additional anti-scam measures across its platforms, including warnings related to suspicious WhatsApp device-linking requests.
Scammers frequently use messaging platforms to impersonate friends, businesses or organisations and persuade people to share sensitive information or send money.
By adding AI-powered detection directly to WhatsApp, Meta hopes to warn users before they engage further with potentially fraudulent conversations.
For now, the Scam Alert feature remains limited to beta testing on Android, but its introduction could provide WhatsApp users with another layer of protection against suspicious messages when the feature eventually reaches the wider public.
WhatsApp Chat With Us Have a question? Send us a message. ›
