[{"data":1,"prerenderedAt":28},["ShallowReactive",2],{"nr-en-google-deepmind-singapore-on-device-scam-detection":3},{"slug":4,"title":5,"dek":6,"date":7,"time":8,"publishedAt":9,"updated":10,"updatedAt":10,"dateFmt":11,"updatedFmt":10,"kind":12,"tier":13,"author":14,"authorName":15,"topics":16,"tracker":10,"trackerLabel":10,"headlineStat":22,"image":23,"ogImage":24,"imageAlt":5,"csv":10,"minutes":25,"words":26,"html":27},"google-deepmind-singapore-on-device-scam-detection","Google DeepMind and Singapore use AI against chat scams – directly on the smartphone","A model developed with the Singapore government detects manipulative conversation patterns locally on the device. According to the sources, message content is not to be sent to a server.","2026-10-14","07:05","2026-10-14T07:05:00+02:00","","October 14, 2026","news","standard","ideal-syka","Ideal Syka",[17,18,19,20,21],"AI safety","Data protection","Fraud prevention","On-device AI","Singapore","Scams: over 50% of reported crimes in Singapore, more than S$4 bn in losses since 2019","\u002Fnewsroom\u002Fimg\u002Fgoogle-deepmind-singapore-on-device-scam-detection.webp","\u002Fog-nr\u002Fgoogle-deepmind-singapore-on-device-scam-detection.en.png",2,330,"\u003Ch2>What was announced\u003C\u002Fh2>\n\u003Cp>According to Google DeepMind, Google DeepMind and the Singaporean agency GovTech have developed a scam detection model meant to run on typical smartphones in Singapore. The base is Google DeepMind's lightweight Gemma model, fine-tuned on local scam conversations. The announcement is dated 14 October 2026 and was presented, according to the blog, at the Singapore International Cyber Week (SICW 2026).\u003C\u002Fp>\n\u003Ch2>Background\u003C\u002Fh2>\n\u003Cp>According to Google DeepMind, scams account for more than half of the crimes reported in Singapore and have caused more than S$4 billion in losses since 2019. Tactics have shifted from simple phishing to social engineering, in which attackers manipulate victims through convincing conversations before requesting money. The source says such scams account for up to 80% of all losses.\u003C\u002Fp>\n\u003Ch2>How the model works\u003C\u002Fh2>\n\u003Cp>According to the source, the model analyses conversations directly on the device. The reason: many messengers are end-to-end encrypted, so server-side scanning is not possible there. When the model detects manipulative patterns, it shows a warning on the device advising caution before sharing sensitive information or funds, according to the announcement.\u003C\u002Fp>\n\u003Cp>The processing is two-stage: a lightweight classifier screens incoming messages with low power consumption; only on suspicion does contextual evaluation by the fine-tuned Gemma model follow. According to the source, training took place within GovTech's managed GCP environment, excluded personal data, and used anonymised datasets from scam reports and intelligence feeds.\u003C\u002Fp>\n\u003Ch2>Assessment\u003C\u002Fh2>\n\u003Cp>For German companies and data protection officers, the example is relevant because it shows protective functions need not necessarily transmit content to servers. Whether a comparable approach can be implemented in the EU does not depend on the technology alone, according to the sources available; it remains open, for example, how messaging providers and operating system vendors would integrate such models. The source names this integration as a possibility but cites no concrete commitments.\u003C\u002Fp>\n\u003Ch2>Sources\u003C\u002Fh2>\n\u003Cul>\n\u003Cli>\u003Ca href=\"https:\u002F\u002Fdeepmind.google\u002Fblog\u002Fon-device-ai-scam-detection-mddi\u002F\">Google DeepMind\u003C\u002Fa>\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Cp>\u003Cem>Editorially owned by \u003Ca href=\"\u002Fen\u002Fautor\u002Fideal-syka\">Ideal Syka\u003C\u002Fa>. Sources and method: \u003Ca href=\"\u002Fen\u002Fredaktion\">Newsroom & method\u003C\u002Fa>. Tips and corrections: \u003Ca href=\"mailto:ai@i6eal.de\">ai@i6eal.de\u003C\u002Fa>.\u003C\u002Fem>\u003C\u002Fp>\n",1791966628932]