Google DeepMind and Singapore partner on fine tuned AI scam detector

Google DeepMind and the Government Technology Agency of Singapore have collaborated to build an on-device artificial intelligence scam detection system. As reported by DeepMind Research News, AI News, and UA.NEWS, the system uses lightweight fine-tuned Gemma models to intercept conversational scams inside private chat applications without transmitting user data to cloud servers. [1][3][4][5]

DeepMind Research News reported that scams such as investment fraud, impersonation of public officials, and fake job offers make up over half of all reported crimes in Singapore. These financial scams have caused more than S$4 billion in total losses since 2019. Tactics have increasingly shifted from simple phishing links to complex social engineering attacks where scammers engage victims in long conversations before requesting funds. [1]

The project relies on Google DeepMind, an artificial intelligence research laboratory formed through the merger of DeepMind and Google Brain under Alphabet Incorporated. The agency works alongside the executive branch of the Singapore government to deploy privacy-preserving tools directly on consumer hardware. [7][8]

According to reports by Google DeepMind and Tech Edition, the deployment of lightweight on-device models aims to address social engineering tactics that now account for up to 80 percent of all financial losses from scams. The system ensures zero cloud data egress to maintain strict user privacy during message analysis. [1][2][6]