WhatsApp Tests AI Scam Alert to Protect Users From Fraud

Meta is testing a new artificial intelligence feature on WhatsApp designed to protect users from fraudulent and scam messages, according to reporting from Malwarebytes. The optional beta feature utilizes an on-device machine-learning model to flag likely scam messages sent by accounts that are not currently stored in a user’s contact list.

How WhatsApp Scam Alert Works on Device

Once enabled by a user, the Scam Alert tool downloads a machine-learning model directly to the device. According to Malwarebytes, the software examines incoming messages from non-contacts for linguistic signals and conversational structure. The model has been trained on patterns associated with scams previously reported by users.

Because the processing happens locally, WhatsApp states that the system adds defensive friction without requiring the platform to read private conversations. Malwarebytes reports that the feature is designed not to block messages outright. Instead, it displays a warning banner inside the chat if the model identifies a potential scam, prompting the user to pause and evaluate the interaction.

User Controls and Privacy Safeguards

When a warning banner appears, the sender does not see any indication that their message has been flagged. According to Malwarebytes, this ensures that scammers are not tipped off that their approach has been detected. Users retain full control over the chat and can take several actions once a warning is displayed.

• Block the sender to prevent any further messages.
• Report the chat directly to WhatsApp.
• Continue the conversation if the user believes the message is legitimate.
• Mark the chat as trusted, which removes the warning and stops Scam Alert from flagging that specific conversation in the future.

Malwarebytes notes that the feature’s optional nature, local classification, transparency commitments, and lack of automatic reporting represent notable design choices for an encrypted messaging service.

Limitations and Current Testing Status

According to Malwarebytes, some of the most effective WhatsApp scams originate from compromised accounts belonging to people the victim already knows, such as recent account-takeover campaigns involving vote requests. Because those messages arrive from trusted contacts, an unknown-sender warning will not appear. The Scam Alert tool is currently part of Meta’s broader anti-scam campaign across WhatsApp, Facebook, and Messenger to address fraudulent tactics like impersonation, fake jobs, romance baiting, and investment fraud. Malwarebytes reports that the feature is in a limited beta rollout, being tested with researchers in Meta’s bug bounty community before any wider public release.