Google DeepMind and Singapore's GovTech collaborate on on-device AI scam detection using Gemma models
Google DeepMind and GovTech Singapore have developed a fine-tuned, lightweight on-device AI model based on Gemma to detect conversational scams in private chats. According to the source, the model runs locally and sends no data to the cloud.
Why it matters
The project shows a pattern for building fraud or scam filtering that runs on the user's device, avoiding cloud data egress. Builders handling customer messages can evaluate whether on-device classification fits their privacy and latency requirements.
Who should care
Messaging app developers, fintech and marketplace platforms, trust and safety teams, mobile developers working with on-device ML
What you can do
Review the DeepMind post to assess whether a small Gemma-based classifier could run on-device for flagging suspicious messages in your app.