Scammers keep finding new ways in. Meta keeps pushing back. The latest move comes from WhatsApp itself. A feature spotted in beta code promises to flag suspicious messages from unknown contacts. It does so without ever sending those conversations off the device.
Android Police first reported the development on June 2, 2026. The tool, dubbed Scam Alert, analyzes incoming texts locally. https://www.androidpolice.com/scam-detection-but-on-whatsapp/ Its description states plainly, “Your messages always stay private and end-to-end encrypted.” Another line reassures users, “no one who messages you can see you’re using Scam Alert.”
That’s no small detail. WhatsApp’s core promise rests on encryption. Any cloud-based scan would break it. Google already offers similar protection inside its own Messages app. But that system-level approach stops at the boundary of third-party programs. WhatsApp had to solve the problem on its own. So it did. On-device processing lets the app inspect patterns that match common fraud scripts while the chat stays sealed.
When a message triggers the alert, users see a clear choice. Block the sender. Or tap Trust and keep talking. The feature won’t decide for you. It simply surfaces the warning. And it ships disabled by default. Anyone who wants it must turn it on manually in settings. The implementation appeared in WhatsApp beta for Android version 2.26.22.2. Widespread availability remains unclear.
But this beta discovery fits a larger pattern. Meta has spent the past year expanding anti-fraud efforts across Facebook, Messenger, and WhatsApp. In March 2026 the company detailed several new protections. https://about.fb.com/news/2026/03/fighting-scammers-protecting-people-with-new-technology-and-partnerships/
One targets device linking on WhatsApp. Scammers often pose as contest organizers or support staff. They persuade victims to enter a phone number on a fake site, then send a linking code or QR code. Accepting it hands control of the account to the attacker. The new warning shows where the request originates and labels it suspicious. Behavioral signals drive the alert. Users get time to stop and think.
TechCrunch covered the rollout the same day Meta announced it. Reporter Aisha Malik noted that scammers “try to avoid detection and may not immediately use accounts maliciously.” https://techcrunch.com/2026/03/11/meta-rolls-out-new-scam-detection-tools-to-facebook-whatsapp-and-messenger/ The company therefore focuses on early friction. Small pauses. Clear information. Enough to break the rhythm of a well-rehearsed con.
Messenger gains expanded scam detection in more countries. When a new chat shows patterns typical of job offers or other classic schemes, the app issues a warning. It then offers to send recent messages for AI review. Detection here involves servers. Those shared snippets lose end-to-end encryption during analysis. Meta presents the trade-off directly. Users choose whether to proceed.
The numbers behind these moves are sobering. In 2025 Meta removed more than 159 million scam ads. Ninety-two percent came down before any user reported them. The company also disabled 10.9 million accounts tied to criminal scam centers on Facebook and Instagram. Joint operations with law enforcement yielded hundreds of thousands more account takedowns and dozens of arrests. One Thailand-focused effort alone helped lead to 21 detentions by local police.
Nathaniel Gleicher, Meta’s global head of counter fraud, described the dual strategy. “Our work to combat scams focuses on strengthening defenses on our platforms to stay ahead of scammers. Just as critical to our strategy is supporting efforts that drive collective action across sectors and industries to stop scams at its source.” He spoke after Meta signed an industry accord against online fraud at a United Nations event in March 2026.
Scam tactics have grown industrialized. Networks operate across borders. They recruit via social media, run fake investment schemes in WhatsApp groups, impersonate celebrities, and hijack accounts through device linking. Romance scams, crypto fraud, fake customer service calls. The list lengthens every quarter. Older adults remain prime targets. So do users in emerging markets where mobile money moves fast and formal banking lags.
Google took an earlier swing at conversational fraud. Its March 2025 announcement introduced AI-powered scam detection inside Messages and the Phone app. That system scans conversations in real time and warns about impersonation or urgent money requests. Yet it cannot reach WhatsApp chats. The encryption barrier is absolute. Hence WhatsApp’s decision to build its own on-device model.
The beta version of Scam Alert suggests Meta learned from that limitation. Local analysis preserves privacy. It also reduces latency. No round trip to servers. The model presumably looks for keywords, urgency cues, requests for codes or payments, requests to click links. Exact signals stay inside the app for now. But the user control element stands out. No automatic blocking. No silent filtering. The recipient decides.
Some limits remain obvious. On-device models can lag behind cloud systems in adaptability. Sophisticated attackers evolve language quickly. A clever rewrite might slip past initial patterns. False positives will annoy some users who receive legitimate messages from new numbers. The default-off setting acknowledges that friction. It also means adoption could stay low unless Meta promotes the feature aggressively.
And the broader fight extends past any single app. Meta joined 10 other companies in the Industry Accord Against Online Scams and Fraud. Signatories include Google, Microsoft, OpenAI, Amazon, and others. The agreement aims to close gaps between platforms. Scammers exploit those gaps. A victim blocked on one service simply moves to another.
Education campaigns run alongside the code. In India, Meta partnered with government agencies and actress Neena Gupta for the Scam Se Bacho effort. Similar programs target Southeast Asia and Latin America. The message stays consistent. Verify before you send money. Don’t share verification codes. Pause when something feels off.
Recent social conversation shows the feature already generating interest. On X, users noted the timing. One post observed that Meta’s AI had been fooled on Instagram the day before the WhatsApp beta leak surfaced. The contrast drew wry commentary. Cloud models can be tricked. Local detection offers a different, perhaps harder-to-game, layer.
Yet no single tool solves the problem. Device linking warnings help. Message scanning helps. Account takedowns help. The real test lies in whether these defenses scale faster than the scams themselves. Criminal networks have turned fraud into a business. Complete with call centers, scripted playbooks, and performance metrics. Tech companies now respond in kind. With AI, behavioral signals, cross-industry pacts, and persistent enforcement.
WhatsApp’s Scam Alert represents one more brick in that wall. Small. Focused. Privacy-first. It won’t stop every fraudster. But it gives billions of users a quiet second opinion before they reply to that urgent plea from an unknown number. In a world where a single clicked link can empty a bank account, that extra moment of doubt carries real value.
Meta has signaled more updates will follow. The company continues to expand advertiser verification. It invests in better impersonation detection. It works with police forces from Thailand to Nigeria. The March 2026 blog post ended on a note of persistence. “Our global work to protect people against scammers is never done.”
Neither is the work of those who prey on trust. The contest continues. One alert at a time.
WebProNews is an iEntry Publication