Google and the Singapore Government are expanding their cybersecurity cooperation through two AI initiatives, introducing tools designed to speed up software vulnerability patching alongside a system being developed to detect potential scams directly on Android devices.
Unveiled during Singapore International Cyber Week 2026, the initiatives involve Singapore’s participation in Google’s Fairwind Programme and a collaboration between Google DeepMind and the Government Technology Agency of Singapore (GovTech Singapore). The two projects address different threats, targeting weaknesses in essential systems on one front and scam attempts involving private conversations on the other.
AI-assisted patching for cyber defenders
Through the Fairwind Programme, cyber defenders gain access to Google’s Gemini 4 Argon model and CodeMender harness. Google describes Gemini 4 Argon as a new frontier model, stating that the tools can help defenders identify vulnerabilities, verify them and prepare software patches within an organisation’s secure cloud environment.
The programme aims to shorten the time needed to address security weaknesses. According to Google, the tools can generate verified, deployment-ready patches in minutes, compared with manual remediation processes that may take weeks. This earlier remediation would allow defenders to address vulnerabilities sooner and reduce the risk of disruption to essential systems and services.
The initiative extends Google’s cybersecurity cooperation with Singapore, building on previous efforts that include a memorandum of understanding with CSA, refreshed in 2025 to strengthen national cyber resilience, and the AI Agents Sandbox, which examines how Computer Use agents operate in real-world settings.
Scam detection designed to keep conversations on devices
In parallel, Google DeepMind and GovTech Singapore are developing an AI model that analyses potential scam attempts directly on Android devices. Designed to examine conversational patterns locally, the system displays a warning when it identifies a suspicious interaction, prompting users to exercise caution before financial loss or credential theft occurs.
The approach is intended to protect private communications without relying on cloud-based message scanning. Traditional threat detection relies on analysing messages on external servers, but many messaging services use end-to-end encryption, preventing private conversations from being inspected in this way. Local analysis allows the proposed model to identify potential scams without sending private chat data to the cloud.
The partners are fine-tuning lightweight Gemma models using examples of local scams, including impersonations of public institutions, banks and local figures, alongside schemes circulating across the region. The training is intended to help distinguish ordinary conversations from the pressure and manipulation commonly used in social engineering scams.
Google says the model will undergo continuous training to adapt to changing scam tactics. The company also reports having demonstrated the ability to detect scams locally in real time without exposing private chat data, adding that details on how users can install and use the model will be shared later.
Mark Chen, Director, Trust and Safety at GovTech Singapore, said: “Fighting scams shouldn’t mean sacrificing personal privacy. Working with Google DeepMind, we’re bringing advanced AI protection directly onto citizens’ devices. This allows us to catch highly targeted, local scam attempts in real time while keeping personal communications completely private”
JJ Geewax, Director, Applied AI at Google DeepMind, described the collaboration as combining DeepMind’s research with GovTech Singapore’s local domain and data expertise to address evolving scam tactics.
Existing fraud safeguards and scam education
The initiatives build on Google’s existing fraud prevention measures, including enhanced protection through Google Play Protect. According to the company, these safeguards have blocked more than 3.3 million attempts to install potentially malicious applications across more than 770,000 devices, while preventing 68,000 unique apps from potentially being misused for financial fraud and scams.
Alongside technical safeguards, Google is updating Be Scam Ready, an interactive game designed to help users recognise common manipulation tactics. The company cited studies finding that 10 minutes of gameplay was associated with an estimated 20% lower overall risk of falling victim to a scam. Players were also 47% less likely to mistake a scam for a legitimate message and 24% less likely to incorrectly identify genuine communications as scams. Google is now launching Be Scam Ready 2.0, which reduces playtime by half while, according to the company, maintaining the same level of scam protection.
