Loading...
WhatsApp's new AI tool can warn users about scam messages
WhatsApp, with its over 3 billion users, is a major target for scams

WhatsApp's new AI tool can warn users about scam messages

Aug 14, 2026
10:19 am

What's the story

Meta is using artificial intelligence (AI) to combat the growing number of scams on its platform. The tech giant has announced a new experimental feature called "Scam Detection" that employs on-device AI models to flag potentially fraudulent conversations. The move comes as part of Meta's ongoing efforts to stay ahead of evolving scam tactics and protect users' personal messages with end-to-end encryption.

Scam prevalence

Scams on WhatsApp

WhatsApp, with its over 3 billion users, is a major target for scams.

The Federal Trade Commission (FTC) reported that victims lost $425 million to scams on WhatsApp alone in 2025.

This was part of the more than $2.1 billion lost to social media scams that year.

The FTC also found that WhatsApp was the second most common channel for fraud in 2025.

Functionality

How scam detection works

When enabled, the Scam Detection feature downloads a machine learning model onto the user's device. This model analyzes incoming messages from non-contacts to see if they match known scam patterns.

The on-device AI model is trained on patterns observed in scam conversations reported by users. It performs probabilistic classification based on conversational structure and linguistic signals to identify potential scams.

ADVERTISEMENT

User empowerment

What happens when a scam is detected?

If a message is flagged as a potential scam, the user will see a warning that isn't visible to the other person. They can then choose to block, report, or continue the conversation.

If they trust a chat that was previously flagged, they can mark it as such and even share the last five messages received with WhatsApp to improve feature accuracy.

ADVERTISEMENT

Privacy assurance

User privacy remains paramount

Meta has stressed that no user data is automatically shared with Meta, WhatsApp, or any other third party during this process.

All inference happens on-device and no message content leaves the user's device for classification.

The company also said that it won't deliver a specific model to a specific user and every model version will be published on a public transparency ledger.

ADVERTISEMENT