Meta has started testing a new optional security feature called Scam Alert to protect WhatsApp users against fraud attempts. The feature, which is available in limited beta, evaluates whether incoming messages are fraud attempts by using the machine learning model running on the device. When the system finds a message suspicious, it shows a warning to the user directly in the chat, and this warning cannot be seen by the other party of the conversation. The user can then block the relevant account, report it to WhatsApp or continue the conversation. Thus, Meta aims to provide an additional layer of security to the user without examining the message contents on its servers.
Alerts displayed by Scam Alert do not constitute a definitive fraud determination. If the user thinks that the system has made an incorrect evaluation, he can mark the relevant chat as reliable. After this process, the existing warning is removed and Scam Alert does not show a warning again for the same chat. In addition, when users mark a chat as trustworthy, they can optionally choose to share the last five messages sent to them via WhatsApp in order to contribute to improving the accuracy of the feature. Thus, while Meta analyzes messages by default without removing them from the device, it also offers a feedback mechanism based on the user’s explicit preference to improve the model.
WhatsApp Scam Alert examines messages on the device
According to the information provided by Meta, when Scam Alert is active, no message content leaves the user’s device for classification purposes. Conversations considered suspicious are not automatically reported to WhatsApp, Meta or any other party. This approach is a remarkable detail in terms of WhatsApp’s end-to-end encryption-based messaging structure, because messages do not need to be processed on a central server for fraud detection. However, how accurately the models running on the device evaluate the context will determine the performance of the system, especially when it comes to unusual expressions used in everyday conversations. Therefore, having the option to mark the chat as reliable against false positive results seems functional for the user.
Scam Alert is not the only protection WhatsApp has added to combat fraud. Earlier this year, Meta also introduced a separate security mechanism that detects signs of fraud in requests to connect WhatsApp accounts to other devices. In account takeover attempts, attackers may try to persuade users to share a connection code or connect a foreign device to their account. The new system, which evaluates message contents, carries protection directly to the chat stage. This makes it possible to provide an earlier warning for requests to send money, long-term communication methods aimed at gaining trust, or suspicious messages that direct the user to take a certain action.
WhatsApp’s user base exceeding 3 billion makes the platform one of the communication channels heavily targeted by fraudsters. In particular, messages asking for money transfer and fraudulent methods called “pig butchering”, where attackers establish a trust relationship with the victim over time, usually followed by fake investment offers, are among the threats encountered in messaging services. According to data from the US Federal Trade Commission (FTC), users reported losses of more than $2.1 billion in frauds on social media in 2025. It was stated that $ 425 million of this was related to fraud cases that took place only through WhatsApp. These figures also explain why automated detection systems are moving directly to messaging apps.
The fact that Scam Alert is optional leaves the user with control over how to use the feature. Once the feature is activated, it can be turned off at any time, and according to Meta, message contents are not sent outside the device during normal classification. However, the warnings provided by such systems should not be expected to prevent all fraud attempts on their own; The language and methods used by attackers can constantly change. While performing on-device analysis is a positive choice in terms of privacy, the frequency of false alerts during the limited beta period and how successfully messages in different languages are evaluated will show the practical value of the feature more clearly. At this stage, Scam Alert does not replace WhatsApp’s existing blocking and reporting tools, but rather offers an additional warning mechanism that draws attention to suspicious conversations before the user makes a decision.
Join Channel