Introduction

Artificial Intelligence (AI) is rapidly transforming the global financial sector, with wide adoption across banking, insurance, and other services. According to a study by KPMG, 71% of surveyed financial firms have deployed AI tools. Singapore exemplifies this trend; its major banks, including DBS, OCBC, and UOB, have integrated AI and Generative AI (GenAI) to enhance productivity. For instance, DBS has deployed more than 1,500 AI models across 370 use cases, generating over SGD 750 million in economic value in 2024. However, this growth introduces significant risks regarding transparency, privacy, and accountability.

Applications of AI in finance in Singapore

AI provides a solution to the financial inclusion trilemma by balancing scale, risk management, and profitability. By lowering operational costs, institutions can serve long-tail customers. Approximately 92% of financial institutions in Singapore use advanced AI to support Know-Your-Customer (KYC) and anti-money laundering (AML) operations. UOB’s AI solution processes over 5,700 alerts monthly with a 96% true positive rate. Furthermore, AI is utilised for fraud detection, such as OCBC’s system for credit card fraud, and for customer engagement via chatbots like GXS Bank’s eBuddy, which reduced call volumes by 20%.

Risks of using AI in finance in Singapore

The use of AI raises transparency concerns, often referred to as the ‘black box’ problem, where internal decision-making is too complex for users to understand. Data privacy risks are also prominent, exemplified by instances where employees inadvertently enter confidential corporate data into public GenAI tools. Accountability remains a challenge, particularly in determining liability when third-party AI models provide inaccurate advice. Additionally, bias and fairness-related risks occur when models are trained on datasets containing historical human prejudices. GenAI introduces unique threats such as hallucinations and deepfakes; for example, a finance director in Singapore was recently tricked into transferring US$499,000 via a fake Zoom meeting.

Regulating AI in finance in Singapore

Singapore adopts a regulator-led, collaborative approach using soft law instruments. The National AI Strategy (NAIS) and NAIS 2.0 outline a vision for ‘AI for the Public Good’. The Monetary Authority of Singapore (MAS) has introduced several key initiatives, including the FEAT Principles (Fairness, Ethics, Accountability, and Transparency) and the Veritas Initiative, which provides assessment methodologies for AIDA (Artificial Intelligence and Data Analytics) systems. Project MindForge, a consortium of banks and technology partners, developed a robust GenAI risk framework. Most recently, in November 2025, MAS published a consultation paper for Guidelines on Artificial Intelligence Risk Management, which adopt a proportionate risk-based approach.

Lessons learned

Singapore’s regulatory strategy is broadly consistent with OECD principles, combining high-level principles with practical implementation measures and regulatory oversight. The report identifies a spectrum of regulatory responses, from laissez-faire to rules-based legislation. Singapore’s approach is characterised by ‘deliberate gradualism’, evolving from voluntary principles in 2018 to industry-led operationalisation and formal guidelines. To enhance supervision, MAS utilises Supervisory Technology (SupTech), such as Natural Language Processing to monitor social media and a prototype that automates the assessment of borrowers’ creditworthiness.

The report suggests areas for improvement, including elevating current guidelines into enforceable codes. It recommends the establishment of a dedicated AI regulatory sandbox to provide a controlled environment for testing innovative systems. To govern third-party risks, the paper proposes a shared responsibility framework that allocates liability based on each party’s level of control over the AI model’s lifecycle. This requires financial institutions to perform comprehensive due diligence on third-party developers and use diverse datasets to verify model reliability before deployment.

Conclusion

Singapore’s case study demonstrates that effective AI governance in finance requires a calibrated, multi-layered regulatory ecosystem. By blending principles-based regulation, industry engagement, and SupTech adoption, MAS seeks to balance risk management with the need to preserve Singapore’s status as a leading financial innovation hub. Responsible governance must continue to evolve alongside the technology it seeks to regulate.