When Michael Jackson sang about the man in the mirror, he was offering a lesson in personal accountability that resonates unexpectedly well in South African boardrooms today. He proclaimed that “If you wanna make the world a better place, take a look at yourself, and then make a change”.
On April 10 communications minister Solly Malatsi in effect handed a mirror to every senior executive when the cabinet-approved National Artificial Intelligence (AI) Policy was published for comment (open until June 10). It’s an ambitious and future-focused document that’s leaving many across business, government and the media understandably nervous and uncertain.
Much discomfort stems from what the policy is not. If you were scanning the document for a tidy checklist, a set of mandated models or a neat maturity framework to hand off to the IT department, you would be frustrated. You would focus on the ways the policy can’t be applied or policed. And while the funding, application and policing discussion are vital for effective legislation, broader considerations are helpful.
The policy is deliberately nonprescriptive, signalling intent rather than dictating action. It offers direction without detailed execution. For those looking for the safety of tick-box compliance, this ambiguity is a gift disguised as a challenge. It is forcing us back to the most fundamental element of business leadership, which is human agency.
By not overregulating how AI should be developed, one clear consequence is that power is with the boardroom, not the state. The message appears to be, “We will not tell you how to innovate, but we will hold you responsible for the path you choose.”
We have entered an era in which responsible AI is no longer a handy descriptor in an annual report but a proxy for general governance maturity. The absence of hard controls in the national policy should not be confused for an absence of responsibility. In fact, the opposite is true, and our conscious use of the seemingly infinite power of AI has never been more relevant.
More experienced boards will have the confidence to forge ahead with strategy in the absence of clear policy and will have more risk tolerance to wait for the law to catch up. But in a constrained economy amid global volatility with daily disruptions, many are understandably hesitant without clearer direction.
Here, the direction signalled by the government through the draft AI policy marks a shift in the burden to us as businesses, innovators and citizens. We are being asked to build, use and direct AI capabilities that are not only high-performing but also capable of withstanding rigorous scrutiny. The human element becomes non-negotiable at this point. Put plainly, if you cannot explain how your AI-driven decisions are made, you simply cannot govern them.
The effective application of the updated King V Code on January 1 provides useful guidance and guardrails for this freedom, and the overarching South African legal framework is explicit. Under the Companies Act, the duties of directors and prescribed officers to exercise reasonable care, skill and diligence expected of a person in their position with the same knowledge and experience cannot be excused by the presence of an algorithm in a workflow.
King V takes this further by advocating for integrated thinking in the creation of value over time. It’s bolstered by an understanding that organisations affect, and are affected by, the health and interactions of our economy, society and environment. Principle 9 applies to information governance with aspirational guidance and builds on foundational principles of ethical and effective leadership.
Principle 9 speaks directly to emerging technologies such as AI and recommends that governing bodies provide strategic direction for the ethical and trustworthy acquisition and use of technology. More than just “are we using good AI?”, directors should be interrogating what AI is in use, why, and how this is being overseen by humans.
A board implementing structured AI oversight is not going the extra mile but demonstrating basic compliance with the principles of King V and resisting the attraction of outsourcing a moral compass to AI. They are also actively reducing their liability exposure, because when an AI outcome is questioned, we are already seeing international examples where “the machine made the decision” is not a good enough defence.
More specific interrogations could include the following: what is the AI’s purpose, and how is it supporting strategy? Are we inadvertently allowing it to lead strategy rather than inform it? What are the safeguards in place to responsibly manage and direct these powerful tools? Do we know what an AI hallucination would be, and what do we do about these? What does our AI use policy say? Do we even have one, or are we relying on employees’ discretion as new use cases pop up?
A board implementing structured AI oversight is not going the extra mile but demonstrating basic compliance with the principles of King V and resisting the attraction of outsourcing a moral compass to AI.
Questioning is critical to ensure good governance prevails and that an ethical, transparent and accountable culture supported by prudent controls is in place, irrespective of how complex the underlying code might be.
Maintaining a nonprescriptive approach to AI governance and regulation will help preserve the space for human independence, autonomy and soul. AI tools may be excellent at pattern recognition, but they are notoriously poor at understanding context, nuance or the long-term social implications of what a plain reading might conclude is a statistically optimal choice.
Figuring out proportionality of oversight to risk is where human intellect, insight and experience will prove irreplaceable. For low-stakes automation such as simple data capture tasks the leash can be long.
For decisions involving sticky, private or nuanced scenarios like credit, employment or personal safety, the human hand must be firmly in control. When we think deeply about how to maintain the essence of our humanity in this digital age, we can’t forget that these tools can be used to amplify our reach — but we can never completely let go.
Leaders looking for guidance on next steps in the absence of an AI manual should start with existing frameworks. Responsible AI is not a new, parallel discipline. It is a practical extension of good governance, enterprise risk management, Popia-aligned data environments and internal assurance processes.
You may be surprised that many of the elements of responsible and ethical AI are already there, and clearer articulation, tighter oversight or guidance on scaling are what’s missing.
The draft AI policy brings with it many questions, but also the possibility of many freedoms to innovate responsibly and consciously. It has never been more urgent for us to maintain our independence and oversight over these tools.
The machines will keep getting smarter, but the mirror will continue to show that our accountability remains — as it always has — stubbornly and beautifully human.
• Singh is a partner and De Muelenaere a senior manager in finance and risk advisory at technology and management consultancy iqbusiness.
Would you like to comment on this article?
Sign up (it's quick and free) or sign in now.
Please read our Comment Policy before commenting.