Scam tactics have moved from simple phishing links to long, convincing conversations. The model was trained on anonymised local examples, such as fake officials, banks and job offers, inside a government-managed cloud environment. On the phone, a low-power filter screens messages first, and the fine-tuned model then checks the wider conversation before showing a warning. Rollout inside apps will be phased and needs the user’s consent.
Why it matters
It shows a practical way to fight scams in encrypted chats without scanning messages on a server: keep the model small enough to run on the phone. If it works in Singapore, other governments could copy the approach, and messaging apps may face pressure to build in similar on-device protection.
Original source
Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection
Google DeepMind
deepmind.google
Summary by Second Order AI. Full reporting is at the original source.