Google DeepMind and GovTech Singapore are exploring local artificial intelligence applications to protect messaging users. According to DeepMind, the organizations demonstrate how teams can use "lightweight on-device Gemma models to intercept conversational scams in private chats." This architecture applies specialized models directly on the client device rather than routing chat evaluation through remote networks.
Data privacy remains central to the technical implementation outlined by the developers. DeepMind highlights that the scam detection system functions "with zero cloud data egress." Consequently, conversations inside private chats can be screened for deceptive behavior without transferring sensitive text to cloud infrastructure.