Google and Singapore Develop AI Tool to Detect Scams on Android

Google DeepMind and Singapore's Government Technology Agency (GovTech) are creating a first-of-its-kind on-device artificial intelligence model to alert Android users to potential scam attempts in real time. The initiative was announced on October 14 on the sidelines of the Singapore International Cyber Week, marking a partnership aimed at protecting users from known scams while keeping personal communications completely private.

The system is specifically designed to understand local context by analyzing text exchanges on messaging applications like WhatsApp and Telegram. It is fine-tuned to catch targeted local scam patterns, such as the impersonation of banks and government officials, without letting any personal data leave the device. Google stated that the model processes conversations directly on the device rather than sending data to external cloud servers, which ensures faster alerts and data privacy. When the AI detects manipulative conversational patterns, it triggers a real-time warning on the user's device to caution them against transferring funds or sharing sensitive information.

During development, the model was trained on anonymized datasets from scam reports and intelligence feeds, and it was stress-tested against synthetic conversations reflecting local speech patterns. According to Google DeepMind Applied AI director JJ Geewax and group product manager Miguel de Andres, neither the Singapore Government nor Google will be able to access user conversations through the deployment, as the model runs entirely on the device with the user's consent.

A release date has not yet been announced, and it remains unclear whether the model will serve as a default feature on the Android operating system. Additionally, Google announced that it will give the Singapore Government access to advanced AI cybersecurity tools to help fix vulnerabilities, aligning with the country's efforts to combat digital threats.