Google DeepMind and GovTech Develop On-Device Scam Detection for Android Chats
Google DeepMind and GovTech Singapore are building an on-device scam detection model on Gemma that screens Android chats without cloud uploads.
Google DeepMind and GovTech Singapore have developed an on-device scam detection model that warns Android users about manipulative conversations without sending private messages to the cloud. Announced on October 14, 2026, during Singapore International Cyber Week, the project builds on Google's lightweight Gemma model and forms part of Google's National AI Partnership with the Singapore Government.
The on-device scam detection system runs in two stages. A lightweight classifier screens incoming messages first. When that filter flags a possible threat, the fine-tuned Gemma model evaluates the wider conversation to judge whether it shows manipulation. All processing stays on the handset, so no server scans encrypted chats.
The model was trained on local scam conversations, including impersonation of public institutions, banks and well-known local figures. That focus matters in Singapore, where Google DeepMind says more than half of reported crimes involve sophisticated social engineering. Scams have cost Singapore more than S$4 billion since 2019.
How On-Device Scam Detection Would Reach Users
The target platform is Android. The model analyses text exchanges in messaging apps and is designed to alert users in real time before they transfer money or share credentials. Google DeepMind describes the goal as identifying threats in real time without compromising personal privacy.
No launch date has been set. Google DeepMind says application integration will be phased and developed alongside the Singapore government. No participating app or public performance data has been released, so the detection rate of the model remains unmeasured in public. For now, the detector is a development project rather than a feature users can switch on.
Why this matters
The approach moves scam screening onto the phone itself, which removes the need to upload chats for server-side analysis. For Android users, the open questions are which messaging apps and handsets will receive it, and when. Singapore's rollout will show whether an on-device classifier can catch impersonation-based scams that rely on context.
Sources
On-Device AI Against Scams: Google DeepMind & MDDI
Related Articles
- Google Adds New AI Scam Detection and Anti-Fraud Tools to Android
- Google Debuts New AI Scam Protection Tools to Fight Mobile Fraud
- Google Targets Outsider Enterprise in Major Legal Strike Against AI Scams
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.