AI protection that keeps privacy front and center
Google DeepMind and Singapore’s Government Technology Agency, GovTech, have announced a new AI tool aimed at helping citizens avoid conversational scams. The system uses lightweight on-device Gemma models to identify suspicious scam behavior directly within private chats.
What makes this especially promising is its privacy-preserving design. Because the AI runs on the device, messages do not need to be sent to cloud servers for analysis, enabling zero cloud data egress while still offering timely protection.
A practical win against real-world scams
Scams increasingly rely on long, persuasive conversations to manipulate victims. By detecting warning signs earlier in the interaction, this tool could help people pause, verify, and avoid financial or personal harm before it happens.
- Public safety: Helps citizens recognize scam attempts in real time.
- Privacy-first AI: Keeps sensitive conversations on the user’s device.
- Scalable deployment: Lightweight models make protection more practical across everyday devices.
This collaboration is a strong example of AI being used for practical civic benefit: protecting people from fraud while respecting the privacy of their personal communications.