AI GUIDE FOR BOARDS IN SINGAPORE
Section V: Organisational Resilience and Capability
Chapter 10: Operationalising Digital Resilience
As AI is increasingly embedded in business workows, digital attack surfaces and operational dependencies are expanding. Boards should therefore strengthen oversight over four key areas: • Infrastructure robustness for AI workloads. • Cyber security for AI systems and data. • Operational resilience when AI fails or misbehaves.
• Ongoing model risk management as conditions change.
Boards must move from high-level discussions about AI adoption towards practical assurance that AI-enabled capabilities can operate safely, reliably and consistently.
10.1 Infrastructure Robustness
even when individual components fail or disruptions occur. • Performance under stress: Stress testing before deployment to ensure that AI systems can interact with other enterprise systems.
AI workloads, including model training, real-time inference and large-scale data processing, place signi cant demands on enterprise infrastructure.
10.1.1 Maintaining operational reliability
Boards should approach the Oversight of AI Infrastructure from the perspective of operational reliability, continuity and the organisation’s ability to support AI at scale. Boards should ensure management has considered: • Capacity planning: Infrastructure should be able to support peak AI demand without aecting core business operations. • High availability: AI-enabled systems should remain continuously operational,
Can AI-enabled operations perform reliably and consistently?
Operational readiness is a key shi from high-level discussion about infrastructure resilience. Boards should focus on whether AI infrastructure is engineered, tested and governed to standards consistent with other mission-critical enterprise systems.
90
Made with FlippingBook Digital Proposal Creator