# Google DeepMind and Singapore Build On-Device Tool to Flag Chat Scams

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/31592
Published: 2026-10-14T06:53:26.000Z
Updated: 2026-10-14T06:53:26.000Z
Section: Technology

> The system screens messages on users’ devices and can warn about manipulative patterns. Its release date, accuracy rates and supported apps have not been provided.

Google DeepMind and Singapore’s Government Technology Agency (GovTech) have developed an on-device artificial intelligence tool to flag scam tactics in private chats, offering warnings without sending message content to the cloud.

A lightweight classifier first screens incoming messages. When it flags a possible threat, a fine-tuned Gemma model examines more of the conversation for manipulative patterns and can issue a warning on the device. The model was fine-tuned on local scam conversations.

The system was unveiled Oct. 14 in Singapore (Oct. 13 ET) at Singapore International Cyber Week. No public release date, detection rate, false-positive rate, supported devices or messaging apps have been provided. No integrations have been identified.

Many messaging services use end-to-end encryption, which limits their ability to inspect private chats on their servers. On-device analysis can allow scam warnings without uploading conversations, though users’ access to the tool depends on device makers or messaging platforms integrating it.

Scams account for more than half of reported crimes in Singapore and have caused more than S$4 billion in losses since 2019. The project’s testing used synthetic dialogues and local corpora, but no results were provided.

“Fighting scams shouldn't mean sacrificing personal privacy. Working with Google DeepMind, we're bringing advanced AI protection directly onto citizens' devices,” Mark Chen, director of Trust and Safety at GovTech Singapore, said.