AI governance addresses bias, fairness, transparency, accountability, and regulatory readiness. The textbook draws on NIST AI RMF, the EU AI Act, and Philippine sectoral guidance.
9.3 AI Governance and Ethics
Frameworks to align with
Framework
Scope
NIST AI RMF
Map / Measure / Manage / Govern
EU AI Act
Risk-tiered obligations
ISO/IEC 42001
AI management system
OECD AI Principles
Values + accountability
BSP guidance
PH financial AI use
DepEd / CHED
PH education AI use
Governance components
Component
Practice
Inventory
All AI systems; purpose, owner, risk tier
Risk assessment
Pre-deploy + on material change
Bias / fairness
Measured per protected attribute
Explainability
Reason codes for high-impact decisions
Human oversight
Defined intervention point
Monitoring
Performance + drift + safety post-deploy
Incident response
AI-specific runbook
Worked example - hiring screening AI
Item
Value
Decision
Screen CVs to a shortlist
Risk tier
High (employment outcome)
Controls
Bias audit; reason codes; human review
Drift
Monthly fairness review
Recourse
Candidates can request human review
Audit
Decisions logged with model + input + rationale
What to avoid
AI for surveillance with no consent or oversight.
Black-box decisions in regulated domains.
Treating accuracy as the only metric (fairness matters too).
Pilot-to-prod gap; pilots get reviewed, production does not.
Mentor’s tip: AI fails differently from regular software. Inventory + risk tiers + bias measurement + human oversight. NIST AI RMF as the spine; EU AI Act for cross-border; BSP guidance if you operate in finance.
Discussion
Loading…