Google与GovTech开发端侧诈骗检测模型
Start here
AI-assisted source synthesis · not an independent investigation.
2026年10月14日,Google DeepMind与新加坡政府科技局(GovTech)宣布开发一款端侧人工智能模型,用于在智能手机对话中检测诈骗。该模型基于轻量级Gemma运行,无需将消息发送至云端;系统先由分类器筛查,再分析更大范围的对话内容,在发现操纵模式后,实时提醒用户谨慎分享敏感信息或进行转账。部署需要用户同意,双方称无法通过模型访问用户对话。目前项目将分阶段推出,但尚未公布上线时间或参与应用。
AI-written from reports · updated 2 days ago
Timeline
Follow the reporting to see the different sides of the story.
- Fintech News SingaporeGoogle DeepMind 与 GovTech 开发端侧 AI 模型检测诈骗
Google DeepMind 与新加坡 GovTech 开发端侧 AI 模型,在智能手机对话中检测诈骗,无需将消息发送至云端。模型基于轻量级 Gemma,先由分类器筛查,再分析更大范围对话,发现操纵模式后实时提醒用户谨慎分享敏感信息或转账。部署需用户同意,双方称无法通过模型访问对话;项目将分阶段推出,尚未公布上线时间或参与应用。
Heat over time
Current heat 3·Peak over the comparable range 10 (Oct 14 18:00)·Change over the comparable range in the last 24 hours -50%
Trends compare only the same entities that were fully and continuously observed, so the range can be narrower than the current heat statistics. Move the pointer or tap the chart to read hourly heat; arrow keys work on a keyboard.