Signals in action: A case study for digital infrastructure reputation transparency BOX 7 The Global Signal Exchange (GSE) is a non-profit clearing house for the real-time, international, cross-sector sharing of scam and fraud threat signals. Established in January 2025 with 40 million threat signals, by October 2025 the GSE had grown to 550 million signals. More than 180 organizations have been onboarded or are in the pipeline, spanning major industry providers, technical and non-profit organizations and, now, government agencies: in September 2025, Singapore’s Government Technology Agency became GSE’s first government member. GSE produces sector analysis that incorporates AI and machine learning tools and individually processes 1 million signals each day to create a dynamic threat score that is adjusted in real time through the community’s feedback. The GSE enables both open and group sharing. Participants report that GSE signals are highly accurate, often unique and therefore immediately actionable. Industry “league tables” are powerful tools to benchmark performance across categories of digital infrastructure providers. For example, GSE’s Top Level Domain (TLD) league tables present the share of reported abuse relative to the size of each TLD provider, allowing comparison among similarly sized registries. The tables show that some providers maintain near-zero abuse rates while others have considerably higher rates – with the highest at 22.78%.68 This creates visible incentives for best practice and public accountability. Action 9 – Implement AI-powered abuse screening during digital infrastructure enrolment and beyond: Worsening levels of fraud and infrastructure abuse require detection capabilities matching their speed and sophistication. Traditional manual reviews or static checks can no longer keep pace with automated scams, large-scale domain abuse or identity spoofing. AI-powered shared services offer a scalable way to strengthen defences at the infrastructure layer by detecting and flagging malicious activity at the point of registration or onboarding. AI-powered systems should be built directly into digital infrastructure enrolment workflows to evaluate registration requests in real time. Such systems – potentially operated by existing consortia managing signal- sharing or abuse-reporting platforms – can analyse linguistic anomalies, infrastructure reuse patterns, behavioural signals, identity inconsistencies and reputation indicators to identify suspicious activity early. Privacy-preserving techniques (such as hashing, differential privacy and secure multiparty computation) can enable cross-provider detection while protecting user data. These systems should flag potential risks for human review, monitor post-enrolment activity for emerging threats and be refined continuously through feedback from shared abuse-intelligence networks. At Microsoft, we’re harnessing AI to stop fraud before it starts – protecting billions of digital interactions, shutting down 28 million fraudulent accounts and blocking $4 billion in fraud attempts in just the last 12 months. Using AI, our Central Fraud and Abuse Risk (CFAR) team’s real-time detection and global partnerships are redefining what safety means in a digital world. Kelly Bissell, Corporate Vice-President, Central Fraud and Product Abuse Risk, Microsoft Fighting Cyber-Enabled Fraud: A Systemic Defence Approach 20

Fighting Cyber-Enabled Fraud 2025