WhatsApp Launches Scam Alert: How the New AI That Flags Scams Works

2026-08-16 · 8 min de leitura · Camila Torres

During the week of August 12-15, 2026, Meta announced a feature called Scam Alert for WhatsApp. It uses artificial intelligence to analyze incoming messages and warn when it spots signs of a scam. Before anyone gets too excited: the feature is in a limited beta test, which means it is not yet available to everyone in every country at the same time.

The announcement matters for a second reason too. In the previous article in this series, we showed how AI scams on WhatsApp started fooling even attentive people. Now Meta itself is responding with a native feature. It is a public acknowledgment that the problem grew big enough for the platform to act. That validates, with a heavyweight source, what we had been warning about this type of scam.

The size of the problem: why WhatsApp needed to respond with AI

To understand Scam Alert, it helps to look at the numbers. In Brazil, data from Febraban indicates that the scam where someone impersonates an acquaintance already accounts for about 28% of reported fraud cases, the second most common type in the country. In 2024 alone, 153,000 people fell for it. A number high enough to stop being an isolated case and become a pattern.

Estimates from the GolpeZero portal put the average loss per victim in account cloning or takeover cases in the thousands of reais. Kaspersky, for its part, places Latin America as one of the regions concentrating the most phishing attacks in the world, with Brazil as the main contributor. Different sources, pointing in the same direction.

Numbers like these explain why the answer came from the platform itself, not just from third-party tools. When a scam reaches this scale, the solution needs to reach everyone, in the same place where conversations happen.

WhatsApp just admitted, in practice, what ZapToMe had been warning about: AI scams became a problem big enough to deserve a native feature.

What Scam Alert is and how it works

Scam Alert is a WhatsApp feature that uses a machine learning model to analyze incoming messages and alert when it identifies signs of a scam. The model runs on the device itself, without relying on the cloud, and only the person receiving the message sees the warning.

That local processing detail matters for privacy. Messages do not have to leave your phone to be evaluated, which lowers the cost of a security feature. It is not the same thing as end-to-end encryption, but it is an architecture choice that makes a difference for people who do not want conversation content feeding cloud systems.

Another trait: only the receiver sees the alert, the sender is not notified. That prevents scammers from noticing they were identified and adapting their approach in real time. The feature is optional, in a limited beta phase, and was built together with WhatsApp's bug bounty community.

What you can do after receiving an alert

Getting an alert does not force you to do anything, but there are four natural paths. Each one fits a different situation.

• Block the sender. Ends the contact right away and is the right move when the scam pattern is already clear.
• Report the account to WhatsApp. Helps the platform act against the number at scale, not just in your chat.
• Keep the conversation going on your own. For people who want to confirm the suspicion before acting, always with extra caution.
• Voluntarily share the last messages. Helps train the model, but it is optional and should be a conscious choice of yours, not a requirement.

Worth reinforcing: none of these options replaces your own judgment. The alert is a signal, not a verdict. Legitimate messages can be flagged by mistake, and scams can slip through. The call is still yours.

A scam that already hit hundreds of thousands of people in a single year is not solved with a single layer of protection.

Scam Alert vs. Suspicious Number Database: why the two complement each other

Scam Alert analyzes the content of the message after it reaches you, based on patterns the model learned. It is a valuable layer of protection, but it has a limit: it only acts when the message is already in front of you.

ZapToMe's Suspicious Number Database works differently. It identifies the number based on real reports from other people who were already targeted by the same scam, often before the message is even opened. You look up the number, see if it has been reported, and decide whether it is worth replying.

The two complement each other because they answer different questions. The native feature covers the what: is the content suspicious? The number database covers the who: has this number done this to other people? Together they form a more complete layer of protection than either alone.

In practice, the flow is simple. Before replying to an unknown number asking for money, look it up in the Suspicious Number Database. If the message already arrived and you are suspicious, Scam Alert reinforces the suspicion. Different tools, complementary guards.

An alert that analyzes the message after it arrives is useful, but it does not replace knowing, before even opening the chat, whether that number has been reported by other people.

What to expect from the next steps of this feature

Scam Alert is still in beta, with tests alongside WhatsApp's bug bounty community. The rollout to more users and countries has no confirmed date. If you have not seen the feature in the app, it is not a bug: it simply has not reached your region yet.

This article will be reviewed and updated as the rollout progresses. New security features usually change in the first months, between detection tweaks and false positive fixes. What works today may be refined soon, and what is beta today may become standard.

The most complete protection today is not choosing between native AI and community, but using both.

Every new security feature is welcome. None of them, alone, is enough.

Frequently asked questions

What is WhatsApp Scam Alert?

It is a feature in testing that uses an AI model running locally on the device to identify messages with signs of a scam and alert the receiver, without notifying the sender. The feature is optional and, during the beta phase, is available only to a subset of users.

Is Scam Alert already available to all users?

No. The feature is in a limited beta, and the rollout date has not been confirmed by Meta yet. Availability varies by region and app version. It is worth following WhatsApp updates and the company's official announcements.

Is the sender of a message flagged by Scam Alert notified?

No. Only the receiver sees the alert. That is a deliberate choice: if scammers knew they were identified, they could adjust their approach and change the pattern of their messages. Silence for the sender protects detection.

Does Scam Alert replace the need to check suspicious numbers?

No. The two solve different problems. Scam Alert analyzes the content of the message after it arrives. A database like ZapToMe's Suspicious Number Database identifies the number itself, based on reports from other users, often before the message is opened. One complements the other.