Before the First AI Lawsuit, Fix the Companies Act
South Africa can amend its company law now, or wait for a court to define directors’ AI duties after the damage is done.

Sign up for my Substack daily AI newsletter here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Artificial intelligence has already entered South Africa’s boardrooms. Boards use it to review financial information, forecast market trends and flag risks before strategic decisions. Yet our company law does not specifically govern how boards may use AI, as Rehana Cassim, professor of company law at the University of South Africa, points out. That leaves directors uncertain about their duties and their liability. We can clarify the law now, deliberately and with consultation. Alternatively, we can wait for the first major AI-related loss and let a court decide after the damage is done. The first option is better for everyone.
CONTEXT AND BACKGROUND
The gap is growing quickly. According to PSG Capital, AI use among South African businesses rose from forty-five per cent to about sixty-seven per cent in a year. Only fifteen per cent of businesses have formal AI governance policies, and about thirty-one per cent of directors say AI is still not on their board agenda.
King V, which applies to financial years beginning on or after 1 January 2026, does address AI. It calls for clear accountability for AI outputs and outcomes, and for “human oversight and override mechanisms that are commensurate with the level of risk.” That is valuable guidance. A governance code, however, is not legislation. It sets expectations, but it cannot settle who is liable when an AI-assisted decision causes a loss.
INSIGHT AND ANALYSIS
Cassim identifies the legal tensions clearly. Directors must exercise independent judgement in the company’s best interests, which is hard to do if they cannot understand how an AI reached its recommendation. The business judgement rule in section 76(4) broadly protects directors who made properly informed, good-faith decisions, but it is unclear whether it extends to directors who relied on AI that later proved wrong. Directors may delegate functions, but they cannot give up ultimate responsibility, and the AI itself cannot be sued.
The international picture shows where uncertainty leads. In the United States, seventy-two per cent of S&P 500 companies name AI as a material risk, yet only about eight per cent of companies disclose any board-level AI oversight. Seventeen AI-related securities class actions were filed in a single year. Uncertainty inside organisations is just as striking. In ISACA’s global poll, one in five professionals did not know who is accountable for AI harm in their organisation.
IMPLICATIONS
For policymakers, Cassim’s proposals offer a sensible starting point. The Companies Act could require human supervision and control of AI used in board decisions, and AI literacy for directors. It could encourage a board member with AI expertise or a designated AI officer, and address insurance. A specialist, however, cannot carry the board’s duty on its behalf, so every director needs enough literacy to question what AI tells them. Above all, the law should say plainly what King V implies: decisions may be informed by AI, but they may not be delegated to autonomous systems.
For directors, the lesson does not need to wait for legislation. Record how AI informed a decision, insist on explanations you can understand, and make sure a named person remains accountable for every material outcome.
CLOSING TAKEAWAY
South Africa has a rare chance to get ahead of a problem rather than react to one. Our governance code already points in the right direction, but codes guide and laws bind. Clear rules on human supervision, director competence and liability would protect shareholders, employees and directors themselves. They would also give boards the confidence to use AI responsibly instead of avoiding it or relying on it blindly. The alternative is to let the first costly failure write the rules for us. I would rather we wrote them first.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net



Comments