A few years ago, AI governance meant managing hallucinations. Organizations built guardrails, required human review, and kept pilots tightly scoped. That problem is now the baseline. The harder challenge is governing agentic AI systems that do not merely generate outputs for humans to evaluate, they execute decisions, trigger workflows, and act. In healthcare, where those actions intersect with patient safety and some of the most sensitive personal data in existence, the accountability stakes are categorically different.

This keynote draws on Netcare’s experience building AI governance infrastructure across Africa’s most technologically advanced hospital group, examined through the converging lens of South Africa’s regulatory obligations and the rapidly evolving international framework landscape. POPIA’s accountability and purpose limitation principles, SAHPRA’s growing reach into clinical AI, and King V’s board-level technology governance expectations are not aspirational, they are enforceable. So are the OECD and UNESCO AI Principles, the NIST AI Risk Management Framework, and an EU AI Act that is already straining to keep pace with agentic systems it was not designed to govern.

The real-world signals are impossible to ignore. A US federal court ruling that government retaliation against an AI company’s safety limits was unconstitutional, Mobile tech giants elevating Trust and Safety to a keynote-level strategic pillar, and a papal encyclical co-launched with an AI company co-founder at the Vatican, the governance conversation has moved simultaneously into the courthouse, the boardroom, the consumer keynote stage, and the cathedral. It is no longer a specialist concern.

This session will make the case that AI governance in healthcare is a core clinical and strategic leadership capability, one that must be built into how AI is designed, procured, and monitored, not retrofitted when something goes wrong.

Learning 1: Regulatory literacy

Attendees will leave with a working understanding of what POPIA’s accountability and purpose limitation principles, SAHPRA’s clinical AI expectations, and King V’s board-level technology governance requirements specifically demand of AI systems in a healthcare context, and a practical gap-assessment lens to apply immediately to their own environments.

Learning 2: Framework evolution

The NIST AI RMF, the EU AI Act, and most internal governance policies were conceived for static, advisory AI models. Attendees will understand where these frameworks break down when AI systems execute decisions autonomously, and what additional controls must be layered in to close that gap.

Learning 3: Leadership operating model

The Bartz v. Anthropic settlement, the Anthropic–DoD dispute, Apple’s WWDC 2026 Trust and Safety pivot, and the Vatican’s AI encyclical collectively illustrate that the governance landscape is shifting faster than annual policy reviews can track. Attendees will learn how to structure AI governance as a continuously updated operating discipline rather than a document that ages on a Sharepoint drive.