AI in the boardroom - the personal liability of the board

Company directors who delegate their fiduciary duties to an AI model or adopt its output as their own decision risk losing their Business Judgement Rule protection.

“The King V Code, the Companies Act and common law all dictate that corporate decisions must be made by humans based on reasonably trusted information or advice,” says Elani Vogel, Senior Forensics Manager at Loxton Forensics.

However, directors are often under pressure to make rapid decisions on complex business problems, and the speed of AI promises to accelerate decision-making significantly.

How can they safely use AI in a way that limits their personal liability while satisfying their legal, professional and ethical obligations?

King V Code

The King V Code of Corporate Governance sets out the obligations of the governing body in respect of data, information and technology in Principle 10.

This includes the use of AI in paragraph 109(c)(i) and (ii).

The Code requires that AI be employed with adherence to the values of ethics, human centricity, accountability, transparency, explainability, security, privacy, fairness and trustworthiness.

In addition, there should be clear accountability for its effects, including subjecting its parts and processes to human oversight and override mechanisms that adequately protect the organisation and its stakeholders.

In short, King V directly addresses responsible AI use and the need for independent oversight by directors.

The Companies Act

The Companies Act clearly defines a director, and their duties and obligations, with various aspects of the legislation aligning with the human centricity and independent oversight principles of King V.

Section76 requires that they must exercise their powers and perform their functions, firstly, in good faith and for a proper purpose, and, secondly, in the best interest of the company and with a degree of care, skill and diligence that may reasonably be expected of anyone in their position.

As such, they cannot rely on AI advice or appoint a so-called “AI Director” to the board to replace independent oversight or their obligation to make decisions using their own discretion and with due care and skill.

Common Law

Common law dealing with a director’s duties existed before the Companies Act, which expands on rather than replaces that foundation.

In the well-known 1980 case of Fisheries Development Corporation of SA v Jorgensson, the judge stated that a director exercising reasonable care would not accept information blindly.

They could accept and be entitled to rely on it but would give it due consideration and exercise their own judgment in light of it.

“Directors cannot be indifferent or mere placeholders on the board, nor may they shelter behind culpable ignorance or a failure to understand a company's affairs,” says Vogel.

Several such cases highlight that these weighty obligations should not be reduced to a mere AI query.

An AI Buffer

A December 2025 article published by the University of Fort Hare agrees that directors cannot delegate their decision-making responsibilities to an AI model.

The authors argue that section 76 of the Companies Act only allows these duties to be delegated to a natural or juristic person, and AI is neither.

Using AI directly for decision-making could result in breaches that override the Business Judgement Rule protection if the requirements of Section 76 (4)(a)(i) to (iii) of the Act are not met.

“This is especially risky if a director does not possess the requisite AI expertise, such as understanding the processes needed to achieve accurate responses or the ability to recognise when the model may be hallucinating,” says Vogel.

However, directors can protect themselves by delegating to human experts who have these skills, understand the workings and limitations of the model’s algorithms, can interpret its outputs, and can validate that they are accurate and reliable.

This creates a divide between the use of AI and the decision-making process of the director, which can satisfy the BJR requirements, provided they evaluate the validated information carefully and arrive at their decision independently.

Conclusion

These facts support the position that directors cannot rely on AI to make decisions for them or use its advice indiscriminately to arrive at decisions.

“If they do use AI, they are still responsible for the final decision and must critically consider any information they obtain from it beforehand,” says Vogel.

If not, they run the risk of personal liability under section 77 of the Companies Act for any act or omission they commit in their capacity as a director.