The local focus matters because generic models often struggle with regional deception techniques. Scams such as investment fraud, fake job offers and impersonation of public officials make up over half of Singapore's reported crimes and have cost more than S$4 billion since 2019. Tactics have also shifted from simple phishing to social engineering, where attackers build convincing conversations before asking for money; such scams now account for up to 80% of all losses.

Unlike traditional threat detection that scans messages in the cloud, the model processes conversation data directly on the device. When deployed, it will trigger a real-time warning if it detects manipulative conversational patterns, advising the user to pause before sharing sensitive information or funds. Because everything runs on-device, neither the Singapore Government nor Google will be able to access users' conversations through its deployment, and the model will only read conversations with the user's consent under the host application's terms of use.

Running a complex classifier on a smartphone raises battery concerns, so the teams rely on a two-stage pipeline, though the announcement does not detail how it works. The collaboration was presented at Singapore International Cyber Week 2026, and rollout within applications will be carefully phased in partnership with the government.