AI Guide For Boards in Singapore

AI GUIDE FOR BOARDS IN SINGAPORE

Section VI: Operationalisation

Step 3: Strengthen data infrastructure

e eectiveness of AI systems depends heavily on the quality, accessibility and governance of underlying data. Boards should oversee the foundational data strategy to ensure long-term viability and compliance. Key considerations include: • Data provenance: Boards should ensure management maintains visibility over data sources, ownership and licensing arrangements, and legal rights governing data use. Organisations operating across multiple jurisdictions should consider the diering regulatory requirements relating to privacy and data residency. • Compliance: Boards should oversee compliance with applicable privacy and personal data regulations, and ensure

that management applies appropriate safeguards when using personal information within AI systems. • Infrastructure sustainability: AI deployment may signicantly increase computing demands. Boards, particularly of listed companies, should understand resource implications and monitor the carbon footprint of high-compute AI projects as part of mandatory climate- related disclosures. Sustainable AI is becoming a core component of ESG reporting.

Reference Chapter 9: Ensuring AI and Digital Resilience Chapter 10: Operationalising Digital Resilience Chapter 11: Talent and Technology

Step 4: Establish risk management guardrails

• Testing: Red-teaming and adversarial assessment can help identify vulnerabilities, security aws, or hallucinations before they are deployed in customer-facing environments. • Acceptable use policies: Boards should approve policies that establish clear boundaries for AI usage, including circumstances where human intervention remains mandatory.

Responsible AI practices should be viewed as part of enterprise risk management. Ethics, in the context of AI, is a functional risk mitigation tool. Areas for oversight include: • Algorithmic bias: Periodic independent audits should be conducted to identify and mitigate bias in AI-driven decision- making. • Explainability: Organisations should prioritise AI systems that provide sucient transparency and explainability, particularly in higher-risk applications. “Black box” models (where decision- making is opaque) should be avoided.

Reference Chapter 7: Board-Led AI Governance Chapter 11: Talent and Technology

106

Made with FlippingBook Digital Proposal Creator