AI has arrived in South Africa’s boardrooms – but company law hasn’t caught up

AI is beginning to influence how boards of directors make decisions in South African companies. Directors use it to do things like assess financial information, predict market trends or identify potential risks before making strategic decisions.

But there is a problem. South African company law does not regulate its use by boards. This matters in relation to the rules governing directors’ duties and responsibilities.

Many countries already regulate AI through binding legislation. For example, the European Union, South Korea and Vietnam adopted comprehensive AI laws. Other countries such as China introduced more targeted AI regulations.

As a company law researcher, in a recent paper I looked at four company law implications of AI use in South African boardrooms:

• whether by consulting AI the board will be regarded as having complied with its fiduciary duties
• whether directors can rely on the business judgment rule in section 76(4) of the Companies Act when AI-assisted decisions result in negative consequences for the company
• whether directors can lawfully delegate their functions to AI, and rely on AI-generated advice without incurring liability for incorrect decisions made by those systems
• who is liable when AI errors cause harm to third parties. In other words, who is legally responsible for the consequences of decisions that AI makes?

My findings show that current South African law provides little guidance on these issues. As AI becomes part of everyday business, this gap is becoming a legal and governance challenge. Clear legal rules are urgently needed to prevent a legal vacuum being created in South African boardrooms.

How AI can assist boards

There are different ways in which boards can use AI to manage companies and make decisions.

The simplest form is assisted AI. It performs administrative tasks while directors retain control over decision-making. It can compile meeting agendas, call meetings and draft reports. This allows directors to focus on strategic decisions.

Augmented AI can analyse large amounts of information, identify trends and predict the likely outcomes of different business decisions. For example, it could analyse data and evaluate the risks of a merger and recommend the best course of action. The AI provides recommendations, but the board still makes the final decision.

The most advanced form is autonomous AI, which can make decisions and carry out tasks without constant human input. Boards could, for example, authorise an AI system to make investment decisions within set parameters and independently monitor markets and execute trades.

AI has important limits

Although AI has advantages in supporting corporate decision-making, my research found that it has limitations too. These may affect the reliability of AI-generated recommendations.

AI systems learn from existing data. If that data is inaccurate or biased, AI’s recommendations could also be biased. This is a concern in South Africa, where historical inequalities remain reflected in many datasets.

A 2024 study by the Centre of Excellence in Financial Services suggested that AI used in South African banking could perpetuate racial bias. If companies rely on it without proper oversight, AI could repeat or worsen existing discrimination patterns.

Another challenge is that many AI systems operate as a “black box”. They may produce sophisticated recommendations, but even their developers may struggle to explain how the AI systems reached a particular conclusion.

A further limitation is that AI lacks human qualities essential for effective board leadership, such as empathy, intuition and emotional intelligence.

Can directors delegate their functions to AI?

South African company law allows directors to delegate certain functions to others. But they cannot relinquish their ultimate responsibility for managing the company. They remain accountable for board decisions.

My research argues that the same principle should apply to AI. Directors should be allowed to delegate their functions to assisted AI and perhaps augmented AI, provided they properly supervise the AI system and do not relinquish their responsibilities.

But delegation to autonomous AI should not be allowed. If directors delegated their functions to it, they would be giving AI decision-making powers and giving up their responsibilities.

Directors cannot simply accept AI recommendations. Company law requires them to exercise independent judgment. They must also act in the company’s best interests. If they can’t understand how AI reached a recommendation, how will they determine whether it is in the company’s best interests?

In my view, directors should therefore have some AI literacy. They do not need to understand the coding in detail. But they should have enough AI literacy to understand how the AI works. This will enable them to properly supervise the delegation of their functions to AI.

Who is liable when AI gets it wrong?

An important legal question is determining responsibility when AI causes harm.

Imagine management uses AI to assist in recommending an investment to the board. The board approves the investment based on that AI-generated recommendation, but the investment results in substantial losses. Who should be held liable?

The AI system itself cannot be sued because it is not a legal person. It cannot own property, pay damages or appear before a court.

Responsibility could potentially fall on the company, the software developer or directors who relied on the AI system. Directors might be held liable if they failed to supervise the AI system properly, but the Companies Act does not clearly define where responsibility begins or ends.

This uncertainty creates risks. Companies need legal certainty before investing in AI. Directors need clear guidance about their duties, while those affected by AI-driven decisions need to know who is accountable if something goes wrong.

Without clear legal rules, disputes are likely to increase as AI becomes more common in corporate decision-making.

South Africa should act now

South Africa has taken some steps. The National Artificial Intelligence Policy Framework promotes ethical, transparent and human-centred AI. The Draft National AI Policy is pending. It was published for public comment in April 2026 but was withdrawn after fictitious AI-generated sources were discovered in the reference list.

The King V Code on Corporate Governance for South Africa 2025 also recommends that boards should ensure that AI is used ethically, transparently, fairly, securely and with clear human oversight.

In my view, these developments are important, but they are not enough. The Companies Act should be updated to reflect AI’s growing role in corporate decision-making.

The Companies Act should require directors to supervise decisions made by AI. Ultimate human control of the AI system should be legally required.

My research argues that provision should be made for directors to rely on assisted and augmented AI, but not autonomous AI, subject to the directors supervising the AI and exercising an independent judgment.

The Companies Act should require directors to have a sufficient level of AI literacy. Boards should appoint at least one member with AI expertise. Companies should be required to appoint a designated AI officer responsible for the safe, ethical and unbiased use of AI.

Companies should also consider taking out insurance against potential liabilities arising from AI-generated decisions.

Source link