Singapore GovTech uses Gemma AI to block scam chats on-device, keeping private data off the cloud
Singapore’s government agency uses lightweight Gemma models to intercept conversational scams in private chats without sending data to the cloud.
Key points
- Google DeepMind collaborated with GovTech Singapore to build a fine-tuned AI model for scam detection.
- The system uses lightweight on-device Gemma models to analyze private chat messages locally.
- The solution intercepts conversational scams with zero cloud data egress, preserving user privacy.
Google DeepMind and GovTech Singapore have announced a collaboration to deploy an AI-based scam detection system. According to first-hand research published by Google DeepMind, the initiative focuses on intercepting conversational scams within private chats. The partnership leverages lightweight on-device Gemma models to perform this analysis locally, ensuring that no user data is sent to the cloud. This approach addresses privacy concerns while targeting fraud in real-time communications.
How does the detection mechanism operate?
The system utilizes fine-tuned Gemma models that run directly on user devices. These lightweight models analyze conversational patterns in private chats to identify potential scams. By processing data locally, the technology avoids cloud data egress entirely. This on-device architecture allows for immediate interception of fraudulent messages without exposing sensitive personal information to external servers or cloud infrastructure.
Which organizations are involved in this deployment?
The project is a joint effort between Google DeepMind and GovTech Singapore. GovTech is the digital agency responsible for Singapore’s government technology infrastructure. Google DeepMind provided the underlying AI technology and expertise to fine-tune the Gemma models for this specific use case. The collaboration aims to protect users from scams through advanced, privacy-preserving artificial intelligence techniques developed by these two entities.
What is the primary benefit of this approach?
The main advantage is the preservation of user privacy through zero cloud data egress. Traditional cloud-based analysis often requires sending chat data to external servers for processing. This on-device solution keeps all data local, preventing exposure while still detecting scams. It allows for real-time protection against conversational fraud without compromising the confidentiality of private communications between users.
Questions to ask your suppliers
Security teams should evaluate whether vendors process sensitive data locally or transmit it to external clouds. Ask suppliers to clarify their data handling policies for AI-driven features, specifically regarding on-device processing capabilities. Determine if the vendor provides transparency into model training data and privacy safeguards. Inform your risk management and legal teams about any third-party AI integrations that access private communications. Ensure contracts specify data residency and processing locations to maintain compliance with internal security standards and regulatory requirements.
Background: Data residency
Data residency is the requirement that digital information remains stored within specific geographic boundaries. These rules exist to satisfy local laws or organizational policies. You must configure cloud storage to prevent data from moving across borders. This control limits where your backups and active records physically sit.
Read the full guide: What Is Data Residency: How Location Rules Protect Your Cloud
What to do and how to stay safe: Google DeepMind
- Review current vendor contracts to verify data processing locations for AI features.
- Confirm that any on-device AI tools do not exfiltrate data to external servers.
- Monitor vendor announcements for updates on privacy-preserving AI technologies.
- Assess internal policies regarding the use of third-party AI in communication tools.
Step-by-step guide: Browser Security Settings for Small Business Teams
General security guidance from the TechSec Feed newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Does the scam detection system send chat data to the cloud?
No, the system uses on-device processing with zero cloud data egress, keeping all data local.
Which AI models are used for this detection?
The system uses lightweight, fine-tuned Gemma models developed by Google DeepMind.
Who is deploying this technology in Singapore?
GovTech Singapore is deploying the technology in collaboration with Google DeepMind.
