Google DeepMind and Singapore's Government Technology Agency (GovTech) collaborate on an AI model to detect scams in real-time. The system, announced at Singapore International Cyber Week, runs directly on user smartphones. It provides warnings about manipulative conversations before sensitive information is shared.
Operating on-device, the model analyzes chats within end-to-end encrypted messaging apps. This avoids sending private data to external servers, preserving user privacy. The technology uses a lightweight version of Google's Gemma model. This model was fine-tuned with anonymized data from Singapore-specific scam patterns.
When the AI detects conversational tactics associated with scams, it triggers an on-device alert, prompting user caution. This initiative aims to combat the rise of social engineering scams, which have accounted for significant financial losses in the region. Developers have not announced a public release date for the feature.