AI GUIDE FOR BOARDS IN SINGAPORE
Section IV: Ethics and Trust
Emerging AI Risks
AI SUPPLY CHAINS
GENERATIVE AI SERVICES
AGENTIC AI SYSTEMS
• Reliance on multiple upstream providers, models, datasets, and infrastructure platforms.
Key risks include:
As AI systems become capable of taking actions autonomously, boards should understand:
Hallucinations and inaccurate outputs
AI may generate plausible but incorrect or misleading information. Data leakage Sensitive or proprietary data may be inadvertently exposed. Unclear model training practices Lack of transparency on data sources used to train models.
Actions agents can perform Understand the scope, limits and potential impact of autonomous actions. Permissions they possess Review data access, system privileges and ability to interact with other systems. Approval mechanisms Ensure appropriate human- in-the-loop controls for high-impact actions. Existence and maintenance of sufficient audit trails Confirm actions are logged, traceable and tamper- resistant.
Upstream Providers
Models & Datasets
Infrastructure Platforms
AI Applications & Services
• Limited visibility into these dependencies can create concentration, resilience, and accountability risks.
Concentration Risk
Resilience Risk
Accountability Risk
Rapid model changes without adequate notification
Frequent updates can impact performance, behaviour and compliance.
8.4 Stakeholder Governance
Responsible AI leadership ensures the organisation has the people, behaviours and governance structures necessary to use those technologies responsibly, safely and eectively. Boards should therefore focus on three priorities: • Talent development: Building sustainable AI capability through talent development and ecosystem partnerships. • Organisational culture: Fostering a culture that promotes responsible and accountable AI use. • Ecosystem governance: Governing third-party AI dependencies with appropriate oversight and risk management.
Organisations that successfully align talent, culture and ecosystem governance will be better positioned to realise AI’s bene ts while maintaining trust, resilience and long- term value creation. Boards may consider the following Board Checklist for Stakeholders :
e long-term success of AI depends as much on people, culture and governance as it does on technology.
80
Made with FlippingBook Digital Proposal Creator