1.1-1.2 | Purpose: MAS supervisory expectations on AI risk management for financial institutions, complementing FEAT and existing MAS instruments, applied on a group basis for locally incorp | 0 |
1.3-1.6 | Definitions: AI as machine-based systems or models deriving outputs from learned premises, excluding rule-based tools; model, system and use case; high-level expectations for all F | 0 |
1.7 | Regularly review the adequacy of AI risk management against the pace of AI developments and address new or accentuated risks | 0 |
1.8-1.12 | Effective dates and the risks the Guidelines address: existing risk categories extended by AI, amplified by Generative AI and AI agents | 0 |
2.1-2.2 | Apply the Guidelines proportionately: establish oversight frameworks, policies and procedures with AI identification, risk materiality assessment and an inventory (Sections 3 and 4 | 0 |
2.3-2.5 | Low-impact AI use may rely on basic AI governance: assess potential impact on the FI and on customers, and keep at minimum senior management accountability, permitted and prohibite | 0 |
3.1-3.3 | Board and senior management ensure existing risk frameworks identify, assess and address AI risk, maintain oversight and risk culture, keep AI use compatible with regulatory requir | 0 |
3.4 | The board or its delegated committee approves and regularly reviews the AI governance approach, addresses AI risk explicitly in the risk appetite framework, sets roles for board an | 0 |
3.5 | Senior management implements and regularly reviews the AI risk frameworks, institutes life cycle strategies and controls, sets roles across business lines and control functions, ru | 0 |
3.6 | Local senior management stays accountable for Singapore considerations when group frameworks are adopted, with visibility, input, timely reports, escalation paths for local issues | 0 |
4.1 | The AI risk management framework includes systems, policies and procedures for identifying, inventorising and assessing the risk materiality of AI, applied consistently, with contr | 0 |
4.10-4.13 | Establish and apply consistently a risk materiality assessment methodology for every AI use case covering inherent and residual materiality against risk appetite before deployment, | 0 |
4.2-4.4 | Identify AI use consistently across all business and functional areas with clear definitions, criteria and processes, covering AI embedded in material third-party services; designa | 0 |
4.5-4.9 | Maintain an accurate AI inventory of use cases, systems and models with update policies, linkages to other inventories, key attributes (purpose, approved scope, model type, data, d | 0 |
5.1-5.2 | Plan and implement controls across the entire AI life cycle with clear roles, reviewed regularly for newer technologies, calibrated to risk materiality, with constraints such as li | 0 |
5.10-5.11 | Third-party AI: onboarding, development and deployment controls adequate to materiality, contractual visibility over AI introduction and changes, testing in the FI's own context wi | 0 |
5.12-5.13 | Select AI systems, models and features against the objectives and risks of the use case, documenting the trade-offs between performance, complexity, fairness and explainability in | 0 |
5.14-5.15 | Evaluate and test each system or model before deployment to reliability and safety levels proportionate to materiality: identify key risks and set measurable thresholds, define and | 0 |
5.16 | Technology and cybersecurity controls for AI systems commensurate with materiality under MAS technology risk requirements: secured environments with hardened configurations, segmen | 0 |
5.17 | Document the AI development process for reproducibility and auditability, in proportion to materiality and in enough detail for an independent reviewer to understand and potentiall | 0 |
5.18-5.21 | Pre-deployment reviews by parties not involved in development, under clear policies with assigned roles and scope and independence proportionate to materiality; formal independent | 0 |
5.22 | Technology and cybersecurity reviews before deployment covering technical implementation, system and network security, resilience, recoverability and operational readiness: secure | 0 |
5.23 | Ongoing monitoring of all deployed AI including third-party AI, proportionate to materiality: defined metrics and tiered thresholds for robustness, stability, data quality and fair | 0 |
5.24 | Re-validate or review AI in the portfolio as informed by materiality, with regular independent re-validation of high-materiality use cases and reviews triggered by findings on inte | 0 |
5.25 | Change management for deployed AI: a framework to judge the significance of changes by nature and materiality, with reviews and re-approval before significant changes and compensat | 0 |
5.26 | Controls for retiring or decommissioning AI that is no longer needed or exceeds risk tolerance, covering dependencies, data retention, secure removal from production and stakeholde | 0 |
5.3 | For high-risk AI use cases, develop, review and test contingency plans with fallback options such as alternative systems or manual processes, and tested activation protocols for an | 0 |
5.4 | Data management controls for AI covering fitness for purpose, representativeness including stressed conditions, data quality monitoring for anomalies, drift and bias, classificatio | 0 |
5.5-5.6 | Controls for transparency and explainability proportionate to the use case and its materiality, with greater attention for credit, insurance, advisory and fund management uses to i | 0 |
5.7-5.8 | Define fair outcomes and control for harmful bias and discriminatory outcomes across the life cycle, with fairness assessments that define protected attributes and proxies, measure | 0 |
5.9 | Human oversight controls proportionate to materiality, accounting for automation bias and decision fatigue, covering roles and escalation, capable and empowered personnel, design t | 0 |
6.1-6.2 | Ensure the competence and conduct of personnel developing, deploying and using AI through recruitment, training and support, allocate human, technological and financial resources p | 0 |
6.3 | Ensure technology infrastructure (processing units, network, memory, secure data pipelines) is adequate for the performance, scalability and resilience needs of each AI use case, s | 0 |