196. FUTURE GOVERNANCE OF AUTONOMOUS UTILITIES

1. Introduction

The future governance of autonomous utilities concerns the legal and regulatory framework for electricity, water, gas and other essential-service providers that increasingly rely on artificial intelligence (AI), machine learning, smart grids, robotics and automated decision-making. An autonomous electricity utility could automatically balance supply and demand, determine network responses, detect faults, manage distributed energy resources, disconnect unsafe equipment and make operational decisions with limited human intervention.

Although automation can improve efficiency, reliability and grid resilience, it creates fundamental legal questions concerning accountability, administrative justice, cybersecurity, transparency, human rights and liability. Autonomous operation cannot mean autonomous legal authority. Public utilities remain subject to constitutional and statutory obligations even where decisions are technically generated by algorithms.

2. Constitutional and Administrative Governance

In South Africa, autonomous public utilities would remain constrained by the Constitution of the Republic of South Africa, 1996, particularly the principles of legality, accountability and administrative justice. Section 33 protects lawful, reasonable and procedurally fair administrative action.

Future regulation should therefore ensure that an automated decision can be attributed to a legally responsible institution. A utility cannot escape responsibility by arguing that an AI system independently produced the decision. Where automated decisions materially affect consumers—for example, disconnection, tariff classification or access to electricity—affected persons should ordinarily have access to reasons, review procedures and meaningful human oversight.

3. Algorithmic Accountability and Transparency

Autonomous utilities will require an algorithmic accountability framework. Regulators may need powers to examine AI models, data sources, decision rules and system performance. Important decisions should generate reliable audit trails showing what information was considered and how the system reached its outcome.

Transparency does not necessarily require publication of every line of proprietary software. Instead, governance can combine protection of legitimate intellectual property with sufficient explainability and regulatory access to determine whether automated decisions are lawful, discriminatory, irrational or technically unsafe.

4. Human Oversight and Liability

A central future principle is likely to be human-in-command governance. Highly automated systems may conduct routine network operations, but critical decisions should remain subject to appropriate human supervision and emergency intervention.

Liability must also remain identifiable. Depending on the circumstances, responsibility could arise for the utility, software developer, equipment manufacturer, system operator or contracted AI provider. Legislation and licences should allocate duties clearly so that automation does not create an accountability gap when consumers suffer loss or essential services fail.

5. Cybersecurity, Data and Resilience

Autonomous utilities will depend heavily on interconnected sensors, smart meters, cloud systems and digital communication networks. This increases exposure to cyberattacks, manipulated data and systemic technological failures. Governance should consequently require cybersecurity-by-design, continuous monitoring, incident reporting, redundancy and emergency shutdown mechanisms.

Personal information collected through smart infrastructure must also be governed consistently with privacy and data-protection requirements. Energy-consumption information can reveal detailed behavioural patterns and should therefore be collected and processed only on a lawful and proportionate basis.

6. Case Law

Joseph v City of Johannesburg 2010 (4) SA 55 (CC)

Facts: Residents challenged the termination of electricity supplied to their building without adequate procedural protection.

Legal Issue: Whether electricity termination constituted administrative action requiring procedural fairness.

Judgment: The Constitutional Court recognised the importance of electricity services and held that the applicants were entitled to procedural fairness before termination.

Legal Principle/Ratio Decidendi: Decisions concerning essential public services can attract administrative-law duties of fairness.

Significance: In an autonomous utility, automated disconnection should not eliminate procedural protections merely because a machine executes or generates the decision.

State Information Technology Agency SOC Ltd v Gijima Holdings (Pty) Ltd 2018 (2) SA 23 (CC)

Facts: SITA sought to challenge its own procurement decision after an ICT contract had been concluded without constitutionally compliant procurement procedures.

Legal Issue: Whether an organ of state challenging its own conduct should proceed under PAJA or the constitutional principle of legality.

Judgment: The Constitutional Court held that the state's challenge to its own exercise of public power proceeded through legality review and declared the procurement decisions constitutionally invalid.

Legal Principle/Ratio Decidendi: The exercise of public power must conform to the Constitution and principle of legality; unlawful public power is constitutionally invalid.

Significance: Future autonomous utilities cannot avoid constitutional control by delegating functions to algorithms or automated platforms.

7. Conclusion

Future governance of autonomous utilities should combine technological autonomy with legal accountability. Effective regulation will require algorithmic transparency, human oversight, cybersecurity, data protection, review mechanisms and clearly allocated liability. AI may increasingly operate essential infrastructure, but ultimate responsibility must remain with identifiable legal actors. The central governance principle is therefore that automation may transform how public utilities make decisions, but it cannot remove those decisions from constitutional and regulatory control.