- Published on
Google DeepMind and Singapore Government announce new AI tool to protect citizens against scams
- Authors
- Name
- DeepHorus
Google DeepMind and Singapore’s Government Technology Agency (GovTech Singapore) have unveiled a fine-tuned, on-device AI model designed to detect scams. The tool, announced under the National AI Partnership, aims to protect citizens without compromising privacy. It was presented at Singapore International Cyber Week 2026.
Scams account for more than half of Singapore’s reported crimes. They have caused over S$4 billion in losses since 2019. Tactics have shifted from simple phishing to complex social engineering, which now drives up to 80% of all losses.
The model uses Google’s lightweight Gemma architecture, fine-tuned on local scam conversations. These include impersonations of public institutions, banks, and local figures, plus regional schemes. This helps distinguish benign chatter from genuine manipulation.
Unlike server-side scanning, the model processes conversation data directly on-device. With user consent, it analyzes patterns and issues real-time warnings before sensitive information or funds are shared. Neither Singapore’s government nor Google can access those conversations.
A two-stage pipeline manages battery efficiency. Stage one screens inbound messages with a lightweight classifier. Stage two uses the fine-tuned Gemma model to analyze a larger conversation snapshot and confirm threats or discard false alarms.
Governance remained strict throughout development. Fine-tuning occurred in GovTech Singapore’s sovereign sandbox on GCP. Training excluded private or personally identifiable information, relying on anonymized scam reports and intelligence feeds. The model was stress-tested against synthetic dialogues and local corpora.
The rollout within applications will be carefully phased with the Singapore government. The project offers a model for messaging platforms balancing user protection with end-to-end encryption. It also shows how governments and AI providers can build technology for public good.
