Google DeepMind and Singapore GovTech debut on-device Gemma scam detector

Google DeepMind and Singapore's Government Technology Agency have built an on-device AI scam-detection tool using a fine-tuned Gemma model that runs on smartphones. The tool uses a two-stage pipeline: a lightweight classifier for initial screening, followed by the fine-tuned Gemma model for contextual threat confirmation. Because processing happens on-device, it can inspect end-to-end encrypted messages that server-side scanning cannot reach, without transmitting conversation data externally. The model was fine-tuned on Singapore-specific scam scenarios including impersonations of public institutions, banks, and regional schemes. Scams account for over half of Singapore's reported crimes and more than S$4 billion in losses since 2019, with social engineering schemes responsible for up to 80% of losses.