Google DeepMind and Singapore Government announce new AI tool to protect citizens against scams
Source note: Primary statements from a frontier lab.
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
Every model that read this
| Model | Provider | Stage | Score | Conf. | Latency | Prompt | When |
|---|---|---|---|---|---|---|---|
| Llama 3.3 70B | Meta | analysis | +50 | 80% | 4264ms | v1.0.0 / m1.0.1 | 2026-10-14 03:17 |
| Claude Sonnet 5 | Anthropic | consensus | +22 | 35% | 5437ms | v1.0.0 / m1.0.1 | 2026-10-14 03:17 |
| Gemini 2.5 Flash | consensus | -100 | 90% | 1575ms | v1.0.0 / m1.0.1 | 2026-10-14 03:17 | |
| Mistral Small 3.1 24B | Mistral AI | consensus | -35 | 80% | 4702ms | v1.0.0 / m1.0.1 | 2026-10-14 03:17 |
AI tool protects citizens against scams.
This is a press-style announcement with no performance data, deployment scale, or independent verification. The stated goal, protecting citizens from scams while preserving privacy via on-device processing, is a modest positive application if it works as described. Lack of evidence limits confidence in the claimed benefit.
This item describes a specific deployment of AI designed to protect citizens from scams. The 'on-device' aspect and 'zero cloud data egress' enhance privacy and security, leading to a positive societal implication.
The system intercepts scams in private chats, which is a significant intrusion into personal communications. The lack of cloud data egress mitigates but does not eliminate privacy concerns.
Evidence extracted
- Google DeepMind and GovTech Singapore use on-device AI to detect scams
SOURCE Google DeepMind (tier 1)
↓
DOCUMENT 528dc152-8918-4c15-a59e-d55b7dd6c516
https://deepmind.google/blog/on-device-ai-scam-detection-mddi
↓
EVIDENCE 1 extracted excerpt
↓
MODEL RUN 4 runs, methodology 1.0.1
↓
SCORE -16 (Mixed or uncertain)
↓
CONFIDENCE 71%