Archive · 14 Aug 2026
WhatsApp announces AI tool to detect scammers on platform
WhatsApp is a popular app that lets people send messages to friends and family.
Sometimes, bad people use the app to try to trick others and steal their money.
WhatsApp has created a new helper that uses artificial intelligence, a smart computer brain that learns patterns.
This helper works right on your own phone, so your messages do not have to leave your device.
When someone you do not know sends you a message that looks like a scam, the helper shows you a warning.
The person who sent the message cannot see the warning.
You can then block the sender, report them, or keep chatting if you want.
If you trust the chat, you can tell WhatsApp, and the warning will not come back.
WhatsApp is testing this feature now, so it might not be available to everyone yet.
WhatsApp says this helps keep people safer while keeping their private messages secret.
WhatsApp announced an experimental feature called Scam Detection that uses on-device AI models to flag suspicious conversations.
The on-device model identifies incoming messages from non-contacts that match patterns observed in scam conversations users have reported to WhatsApp.
Warnings triggered by the feature are visible only to the user and will not be shown to the person on the other side of the conversation.
Users can block, report, or continue a flagged chat, and can mark a chat as trusted to remove the warning permanently.
Per US FTC data cited via The Verge, WhatsApp was the second-highest source of fraud contact methods in 2025.
- Who
- WhatsApp, the messaging platform owned by Meta
- What
- Announced an experimental on-device AI feature called Scam Detection that flags suspicious conversations from non-contacts
- When
- Thursday (exact date not specified in the articles)
- Where
- On users' devices via the WhatsApp app, as announced in a Meta blog post
- Why
- To combat rising scams on the platform, including impersonation, social engineering, and AI-generated lures, while protecting personal messages with end-to-end encryption
- Feature name
- Scam Detection (also referred to as Scam Alert)
- Status
- Experimental
- How it works
- A machine learning model is downloaded to the user's device and performs probabilistic classification of incoming non-contact messages using conversational structure and linguistic signals
- Training data
- Patterns observed in scam conversations from reports users have sent to WhatsApp
- User options on flagged chats
- Block the person, report them, continue the conversation, or mark the chat as trusted
- Opt-in data sharing
- Users who trust a chat can opt in to share the last 5 messages received with WhatsApp to improve the feature's accuracy
- Privacy
- No message content leaves the device for classification; no user data is automatically shared with Meta or third parties
- 2025 fraud figures
- Per US FTC data via The Verge, WhatsApp was the second-highest source of fraud contact methods in 2025; the article cites $4.25 billion in WhatsApp-linked fraud within a $2.1 billion social media fraud total (figures as reported, internally inconsistent)
