Overview
- GovTech and Google DeepMind announced the collaboration at Singapore International Cyber Week on Wednesday, Oct. 14, 2026, to develop lightweight, on-device models based on Google’s Gemma family.
- The tool will analyse private chats locally with zero cloud data egress, a design choice the government says is meant to preserve user privacy and avoid trust concerns about government access to messages.
- Officials gave no rollout date or published performance figures, and success will depend on cooperation from messaging platforms and device makers plus ongoing model updates to match evolving scam tactics.
- The new on-device model is intended to join Singapore’s existing anti-scam stack, which includes SATIS and rMSE that scan hundreds of thousands of sites daily and disrupt roughly 20,000 malicious sites each month.
- Separately, the Cyber Security Agency issued quantum-safe migration guidance for critical infrastructure by 2031 and broader sectors by 2035, and announced a time‑bound kickstarter grant running Nov. 1, 2026 to Oct. 31, 2027 to fund pilots.