Protection that runs on-device
Google DeepMind and GovTech Singapore, the city-state's public technology arm, have announced a new AI tool designed to protect citizens from conversational scams. The tool runs lightweight Gemma models directly on the device.
No data leaves the device
What stands out is that private chats are never sent to the cloud. Because detection happens entirely on-device, message contents are not transmitted to servers, offering a significant privacy advantage.
Why it matters
Scam messages often begin inside a user's own private chat apps. Scanning that content in the cloud raises privacy concerns, but lightweight on-device models change the equation. Google DeepMind's Gemma family serves as an example for this kind of local, privacy-friendly application.
