“The South African Bureau of Standards (SABS) has proposed eight draft artificial intelligence standards covering AI transparency, data quality, human oversight, practical applications and environmental impacts, using existing international ISO/IEC standards as their foundation. The drafts are voluntary at present, but businesses, technology developers and other interested organisations have until 10 November 2026 to submit comments that could help shape South Africa’s approach to responsible AI adoption.”
South Africa’s proposed AI standards signal a new phase in business policy and corporate accountability
Introduction: A policy development with implications for South African businesses
South Africa’s artificial intelligence (AI) sector is entering an important phase as the South African Bureau of Standards (SABS) seeks public input on draft standards designed to guide the responsible development and use of AI technologies. The proposals address transparency, data quality, human oversight, practical applications and environmental sustainability, reflecting growing international attention to the governance of automated systems.
For businesses, the development matters because AI is increasingly used in customer service, marketing, financial services, recruitment, logistics, manufacturing and decision-making. Companies are adopting technologies that can improve productivity, reduce operating costs and support innovation. However, these benefits also bring questions about privacy, reliability, discrimination, accountability and the ability of people to understand or challenge automated decisions.
The draft standards offer a common reference point for organisations seeking to manage these risks. According to reporting by Polity and MyBroadband, the proposed documents adopt existing international standards developed by the International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC), without local modifications.
The immediate priority is consultation. Interested organisations have until 10 November 2026 to submit comments to the SABS. Although the drafts are not automatically legally binding, they could influence business practices, procurement requirements and expectations around responsible AI.
Eight draft documents address key AI governance issues
The SABS proposals cover eight draft documents across several interconnected areas. These include SANS 12792 on AI transparency; four parts of SANS 2620 addressing data quality; SATS 8200 on the controllability of automated AI systems; and technical reports covering AI use cases and environmental sustainability.
Together, these documents aim to provide organisations with structured guidance for understanding, deploying and managing AI systems.
The transparency proposals are particularly relevant to companies that use AI to interact with customers or support important business decisions. Clear information about a system’s intended purpose, limitations, underlying model and supporting data can help organisations understand what a system can reasonably achieve.
For example, a retailer using AI to recommend products should understand which customer information informs its recommendations and whether the system performs differently across customer groups. Similarly, a financial services company using automated tools should assess whether the information supporting its models is appropriate for the decisions being made.
Transparency does not necessarily mean publishing confidential source code or proprietary commercial information. Rather, the proposed framework provides categories of information that stakeholders can consider disclosing, depending on the system and circumstances.
The data-quality documents address another fundamental issue: AI systems depend heavily on the quality of the information used to develop and operate them. Inaccurate, incomplete, outdated or unrepresentative data can undermine performance, even when the technology itself functions as designed.
For businesses, this means that data governance should become part of AI planning rather than an afterthought. Companies need to understand where their information originates, how it is maintained and whether it is suitable for its intended purpose.
Human oversight remains essential
The proposed SATS 8200 document focuses on the controllability of automated AI systems. Its purpose is to provide a framework for monitoring a system’s state and enabling an appropriate controller to intervene when necessary.
This issue is particularly important when AI influences decisions affecting employment, access to services, customer eligibility or financial outcomes.
Automation can accelerate processes, but organisations remain responsible for deciding how their systems are deployed and supervised. A business should therefore establish who can intervene when an AI system produces an unexpected result, how problems are escalated and how decisions can be reviewed.
Human oversight also needs to be meaningful. An employee who cannot understand a system’s output, access relevant information or reverse an inappropriate decision may be unable to provide effective supervision.
Consequently, businesses should evaluate not only the accuracy of AI tools but also the procedures surrounding them. Staff training, escalation arrangements, record-keeping and clear accountability can help organisations identify problems before they cause substantial harm.
International standards could strengthen business competitiveness
The SABS approach draws on internationally developed standards rather than creating an entirely separate set of technical documents for South Africa. This could help local companies align their practices with expectations used in international markets.
For exporters, technology providers and businesses participating in global supply chains, consistent standards can make it easier to explain governance practices to customers, investors and commercial partners.
Internationally active companies may already encounter questions about data quality, system reliability, AI risk management and human intervention. A recognisable standards framework could help South African organisations answer those questions using a common vocabulary.
Smaller businesses may also benefit. Many do not have the resources to develop comprehensive AI governance frameworks independently. A shared reference can help them identify relevant risks and develop proportionate controls.
Nevertheless, adopting international standards does not automatically guarantee that an AI system is safe, accurate or lawful. Businesses still need to consider their own operating conditions, customers, data and legal obligations.
The proposals should therefore be viewed as a potential foundation for better management rather than a substitute for careful implementation.
Voluntary standards are not the same as legislation
An important distinction is that the draft standards are voluntary at present. Their publication does not, by itself, create a new general legal obligation for every South African business using AI. Standards can acquire compulsory effect where the relevant legal mechanisms make them binding, but that is a separate process.
This distinction matters for compliance planning. Businesses should not assume that the consultation creates an immediate new AI licensing regime or that every organisation must obtain certification under the proposed documents.
At the same time, voluntary standards can influence commercial expectations. Customers may ask suppliers to demonstrate responsible AI practices, while procurement teams may use standards to evaluate technology providers. Insurers, lenders and business partners may also consider governance procedures when assessing operational risks.
Companies should therefore monitor the proposals without confusing voluntary guidance with existing statutory requirements.
South Africa’s broader regulatory environment remains relevant. The Protection of Personal Information Act (POPIA), for example, includes provisions concerning certain decisions based solely on automated processing of personal information. Businesses using AI must consider the applicable data-protection requirements regardless of whether they adopt the new standards.
The wider national AI policy challenge
The SABS consultation is taking place while South Africa works on its broader national AI policy. The government withdrew an earlier draft in 2026 after concerns emerged about fabricated references in the document. The withdrawal highlighted the importance of verification, reliable evidence and accountability when AI is used in public policy development.
The standards consultation and the national policy process are related in subject matter, but they are distinct initiatives. The draft standards provide technical guidance based on international documents; a national AI policy would address wider government priorities and the direction of AI development and governance.
A clear relationship between these initiatives could help businesses understand how voluntary technical standards fit into future policy decisions. However, organisations should not assume that the proposed standards will automatically be incorporated into the final national policy.
For policymakers, the challenge is to encourage investment and innovation while addressing risks to privacy, fairness, safety and public trust. For businesses, uncertainty makes it especially useful to develop governance systems that can adapt as the regulatory environment evolves.
Environmental sustainability is also part of the discussion
The draft technical report on environmental sustainability broadens the conversation beyond immediate operational risks. AI systems can have environmental impacts through energy consumption, water use, computing infrastructure and electronic waste across their life cycles.
These considerations are increasingly relevant to businesses assessing the long-term costs of digital transformation. AI applications may deliver productivity benefits, but organisations should also consider the infrastructure required to operate them and whether those resources are being used efficiently.
Companies with sustainability commitments may find it useful to assess the environmental footprint of their technology choices alongside performance, security and cost.
The proposed report provides a reference for examining these questions. It does not, on the available evidence, establish a new mandatory environmental reporting regime for every business deploying AI.
What businesses should do before 10 November
The consultation provides an opportunity for companies, industry associations, technology developers and other stakeholders to raise practical concerns before the standards are finalised.
Businesses that want to participate should review the relevant drafts and follow the SABS commenting instructions. The published guidance identifies dsscomments@sabs.co.za as the submission email address and calls for comments using the SABS commenting template.
Organisations can also use the consultation period to review their own AI practices. A useful starting point is to identify where AI is being used, including through third-party software and service providers.
Companies should then determine whether those systems process personal information, influence important decisions or depend on data that may be incomplete or unreliable. They can review how decisions are explained, who is responsible for supervising automated processes and what procedures exist when something goes wrong.
Businesses with experience implementing AI may be well placed to submit practical feedback on whether the proposed standards are understandable, proportionate and suitable for local operating conditions.
Industry bodies can help smaller firms participate by sharing information about the consultation and identifying common concerns. Broad participation would help ensure that the standards are considered from the perspectives of large enterprises, smaller businesses, developers and organisations adopting AI for the first time.
Conclusion: Preparing for more structured AI governance
The SABS draft AI standards represent an important development in South Africa’s business-policy environment. Their focus on transparency, data quality, controllability and sustainability reflects the need to manage AI as both a commercial opportunity and an area of operational risk.
For companies, the immediate message is not that a new universal AI law has already taken effect. Rather, the consultation offers an opportunity to influence the development of common standards and prepare for more structured expectations around responsible AI.
Businesses that invest in reliable data, effective human oversight, clear accountability and appropriate risk controls may be better positioned to earn customer confidence and work with partners that expect demonstrable governance practices.
The consultation deadline of 10 November 2026 gives interested organisations a defined opportunity to contribute. What follows will depend on the consultation process, the final status of the documents and any subsequent legal or policy measures.
Ultimately, South Africa’s challenge is to create an environment in which businesses can use AI to improve productivity and competitiveness while maintaining accountability, protecting people and building public trust. The proposed standards are one step towards that objective, but their longer-term value will depend on how effectively organisations and policymakers translate guidance into responsible practice.
