“The EU AI Act asks a simple question of anyone building high-risk AI systems: can your organisation reliably produce safe, lawful systems, and can it prove this? A quality management system is how you answer.”

Introduction

Good outcomes stem from good processes and quality cannot be tested into a product at the end; it has to be built in from the start. This seems to be ethos behind the EU’s AI Act (the AI Act) and how we ought to consider artificial intelligence (AI) systems in future.

What does this mean for providers of high-risk AI systems? Well, the organisations that develop high-risk AI systems and place them on the market must now prioritise a quality management system (QMS). A QMS is a written set of policies, procedures and instructions showing how the organisation designs, tests, monitors and fixes its AI systems.

None of this is entirely new. Many organisations already run quality systems under ISO 9001. Privacy law also takes the same approach. The GDPR and POPIA both rest on accountability, this demands that organisations must take responsibility for compliance and be able to prove it. An AI QMS builds on these foundations rather than replacing them.

The AI Act also reaches beyond Europe. It applies to providers outside the EU whose high-risk systems are sold in the EU, or whose outputs are used there. South African developers with European clients must therefore take note and stay alert.

The EU AI Act’s QMS Requirements

The law wants organisations, not just products, to be trustworthy. One compliant AI system proves little, what matters most is whether an organisation has a dependable way of producing compliant systems every time.

The QMS duty sits in Article 17 of the AI Act. It is often confused as a duty to manage risk, but risk management is only one part of the wider system. A QMS must cover the whole life of an AI system: a plan for meeting the rules, how the system is designed and tested, how data is collected and handled, how risks are identified and reduced, and who is responsible for each step.

Through the QMS, an organisation must also show that it has satisfied the AI Act’s other core requirements. These include good-quality, representative data; clear technical records; automatic logs of how the system operates; meaningful human oversight; and accuracy, robustness and  cybersecurity. Once an AI system is in use, an organisation must constantly assess its performance (known as post-market monitoring) and report serious incidents to the relevant  authorities.

How does the AI Act align with the GDPR and POPIA? The requirements in the AI Act mirror familiar privacy principles embedded in POPIA and the GDPR. For instance, good data governance reflects purpose limitation and data minimisation, this translate to the idea that data must only be used for a defined purpose, and should not be collected unnecessarily. From a POPIA perspective, the AI Act expresses the same ideas through its conditions of processing limitation, purpose specification and information quality. Technical records and logs provides for accountability under both laws, meaning that an organisation will be able to show, not merely say, that it complied.

What a Compliant AI QMS Looks Like

A good QMS is a living cycle, not a file on a shelf.  It follows a simple loop: plan what you will do, do it, check whether it worked, and act on what you find. In this essence, the implementation of an OMS must be integrated throughout an organisation and not be treated as another compliance tick-box exercise.

ISO/IEC 42001, an international standard for managing AI across an organisation, adds an AI-specific layer and can be certified. Although certification serves as useful evidence of compliance, it does not by itself prove compliance with the AI Act.

The standard most closely tied to the AI Act is EN 18286, a European standard written specifically for the AI Act’s QMS duty. The EU published its reference in the Official Journal on 19 October 2026. Following a standard published in this way (a “harmonised standard”) creates a legal presumption that the requirements of the AI Act are fulfilled.

A third standard, ISO/IEC 27090:2026, published on 14 October 2026, addresses a different question: how to protect AI systems from attack. It gives guidance on threats such as data poisoning (corrupting training data), model theft and prompt injection (tricking a system with hidden instructions). At this stage, it merely provides guidance and also serves as a useful reference on how organisations can accede to the AI Act’s cybersecurity requirement.

The GDPR requires data protection by design: building privacy into systems from the outset. POPIA requires responsible parties to ensure its conditions are met both when they decide how processing will work and while it takes place. It therefore important to understand that a well-built AI QMS ensures compliance with the GDPR and POPIA if done correctly.

Legal and Regulatory Implications

Rules without consequences do not change behaviour. The AI Act therefore gives regulators real powers, this shifts the spotlight on organisations as they will be forced to prove that they acted responsibly should they find themselves in hot waters with regulators.

Failing to run a proper QMS can lead to fines of up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher.

Fines are not the only risk. Regulators can require changes, restrict sales, or order a system to be withdrawn or recalled, meaning that a system cannot stay on the market if it is a commercial problem, well before it becomes a financial one.

Regulators will ask three questions. Is there a written procedure for each requirement? Is it followed? Can the records prove it? After an incident, the question will focus less on whether the system made a mistake and more on whether the organisation had a way of noticing and if it was reported to the relevant authorities.

South Africa’s Information Regulator follows a similar logic. It can issue enforcement notices and, where these are ignored, impose administrative fines of up to R10 million. Enforcement is increasing.  The Regulator has also begun asking organisations to demonstrate compliance before any complaint is made, and is considering proposals for immediate fines similar to the GDPR model. The message is the same as under the AI Act: be able to prove what you have done.

Specific Considerations

AI rules do not replace existing rules; they sit alongside them. The aim is to fit the AI QMS into the existing governance and not build a separate structure. Therefore the AI Act must be read in harmony with the GDPR and POPIA, for local businesses.

Data Privacy and Cybersecurity

The AI Act’s data and cybersecurity requirements overlap heavily with existing privacy law. Under the GDPR, organisations must assess high-risk processing in advance, build in protection by design, and protect people from decisions made solely by automated means. POPIA has a close equivalent. People may not be subject to a decision with legal or substantial effects based solely on automated profiling unless an exception applies. Where one does, they must be able to make representations and be given enough information about the logic of the processing to do so.

POPIA also requires appropriate security safeguards, notification of security compromises to the Information Regulator and affected people, and a personal information impact assessment. The EU’s NIS2 cybersecurity rules add further duties, and ISO/IEC 27090:2026 offers a shared reference for AI-specific threats. A poisoned training dataset can be a security incident, a data quality failure, and possibly a personal data breach, so one well-designed assessment can often cover all three regimes.

But what are the practical implications of this AI Act?

Organisations that cope best with new regulations tend to share three habits. They start early, they build on what they already have, and they keep the evidence to maintain a record. Fortunately, the AI Act caters for all three habits.

Any programme must begin with the correct timetable. The Digital Omnibus on AI moved the QMS deadlines, which will apply from 2 December 2027 for stand-alone high-risk systems and from 2 August 2028 for high-risk AI built into regulated products. These dates are fixed and do not depend on standards being ready. It is equally important to remember what has not moved: the bans on prohibited AI practices, these have applied since February 2025, and the penalty rules have been enforce since August 2025.

With the timetable settled, the first practical question is : what does an organisation actually have. Very few organisations keep a complete list of the AI systems they develop or use, and fewer still have assessed which are high-risk. This exercise matters as it informs the an organisation’s obligations. A provider carries the QMS duty while a user of someone else’s system carries lighter obligations. However, the line can shift as an organisation undergoes significant changes or rebranding, a system can turn its user into its provider. An inventory that records both risk classification and role is therefore the foundation on which everything else rests.

The next step is to measure existing governance against the Act’s requirements, rather than starting afresh. Most regulated organisations already run quality, risk and privacy frameworks, and the gap is often narrower than it first appears. EN 18286 provides the natural benchmark because it was written for the Act’s QMS duty, and any decision to depart from it should be recorded with reasons that are well-substantiated. ISO/IEC 27090:2026 can serve as a reference point for AI-specific security controls, however, its role as guideline should not be understood as a route to presumed compliance. Privacy compliance belongs in the same exercise and AI assessments can often be combined with GDPR data protection impact assessments and POPIA personal information impact assessments. Moreover involving data protection officers or POPIA information officers from the outset avoids parallel processes that reach inconsistent conclusions.

A system is only as strong as the people accountable for it. Senior management should hold clear responsibility for the QMS, with named owners for each part, so regulators and notified bodies can see who answers for what.

The final consideration is to look outward, to incidents and to suppliers. Monitoring and incident-reporting plans are of little value if they are first tested during a real incident, so organisations should rehearse them in advance, including the timelines for notifying regulators under both the AI Act and POPIA. Much AI is also bought rather than built, which means compliance depends on the supply chain. Contracts should therefore pass obligations down to suppliers through warranties, audit rights and access to documentation. An organisation cannot evidence a QMS for a system it cannot see inside.

Conclusion

The AI Act asks an organisational question before a technical one: can this organisation reliably produce trustworthy AI output, and can it prove this? The QMS is where that question is answered. The underlying ideas will be familiar to anyone who works with the GDPR or POPIA: take responsibility, design a compliance system and keep the records to prove it.

What this means for South African companies

The AI Act is not South African law, but it will most certainly shape South African practice Three groups will be most affected. The first is South African developers that sell or license high-risk AI to customers in the EU. The second is companies whose AI outputs are used in the EU, even where the system itself never leaves South Africa. The third is local subsidiaries of multinational groups, who are usually expected to follow group-wide AI governance built to EU requirements.

The central lesson here is that AI governance has become an organisational discipline, not  just a technical add-on.  The QMS model requires written procedures, named owners, and records that prove compliances with the model. Standards give that discipline its shape and provide guidance on how to incorporate AI governance within an organisation. EN 18286 is a European instrument, but ISO/IEC 42001 and ISO/IEC 27090:2026 are international and can be adopted anywhere.

South Africa is not starting from nothing. POPIA already requires accountability, purpose specification, information quality and security safeguards, and restricts decisions based solely on automated profiling. The Information Regulator is enforcing it with growing confidence. Banks, insurers and energy operators also work within established governance and risk regimes. The gap is one of scope. However, POPIA protects personal information; it does not regulate AI’s safety, robustness, or human oversight.

Nonetheless, for local businesses with European clientele, early alignment is a commercial advantage for those businesses. European customers will expect suppliers to evidence an AI QMS, and procurement contracts are likely to demand it well before any South African law does. Companies that can show alignment will find it easier to win and keep attracting business from that part of the world. They will also be better placed when domestic policy arrives, since it is likely to draw on the same international standards that are already established.

The post AI Quality Management under the EU AI Act: Principles, Obligations and the Links to GDPR and POPIA appeared first on Werksmans Attorneys.

In Part 1, “Confidence Is Not Capability (Part 1): What the Data Shows, and Why POPIA Is the Foundation of Cyber Resilience”, we reviewed the findings of the Data Confidence & AI Readiness in Africa Report 2026, published by Veeam Software together with ITWeb. We argued that POPIA’s security safeguard, data minimisation, breach notification and operator provisions already amount to a cyber-resilience framework, and that data visibility is the foundation on which all of them rest.

This article addresses the technology that will test that foundation most. Only 29% of organisations consider themselves advanced or fully ready to adopt, govern and scale AI using trusted data. [1] Yet AI is already embedded in core operations and autonomous workflows at 14% of them. [2]

The role of AI in cybersecurity: force multiplier and new attack surface

AI as a defender

AI is well suited to the problems the survey identifies. Detection based on machine learning can establish baselines of normal user, system and data behaviour and flag anomalies far faster than human analysts working through alerts. Typical anomalies include unusual access to sensitive datasets, mass encryption activity and abnormal data egress. Automated tools can scan backups for indicators of compromise, verify that restore points are clean, and run recovery tests routinely rather than annually. That directly addresses the gap between recovery confidence and proof of recovery. [3] The report itself highlights automated backup and recovery verification as a means of maintaining audit-ready evidence of resilience, though that passage describes the vendor’s own platform. [4]

AI can also serve as a compliance tool. Automated data discovery and classification can locate personal and special personal information across sprawling areas in the business. That tackles the classification (28%) and visibility (26%) weaknesses respondents report. [5] It also supports the data mapping on which section 19 risk assessments, section 14 retention schedules and section 22 incident scoping depend. Only 15% of respondents identify privacy automation as an investment priority. [6] Yet it can help enforce retention rules, manage data subject requests, monitor cross-border flows and generate the documentation a regulator would expect. For organisations constrained by skills shortages, the single largest barrier at 32%, [7] well-governed AI tooling can extend the reach of small security and privacy teams.

AI as a risk: shadow AI and data leakage

 The same technology creates new exposure. The 11% of organisations reporting informal or employee-led use of public AI tools with limited oversight, [8] and the 17% citing shadow IT or shadow AI as a barrier to data confidence, [9] show the risk of personal and confidential information entering tools outside the organisation’s control. That may amount to processing without a lawful basis, further processing incompatible with the original purpose, an unauthorised cross-border transfer under section 72, or a failure of section 19 safeguards. Where AI systems are used to make decisions about individuals, section 71’s restrictions on decisions based solely on automated processing may also apply. AI also strengthens attackers, enabling more convincing phishing, deepfake-enabled fraud and faster reconnaissance.

The rogue AI agent threat

The most significant emerging risk is agentic AI: systems that do more than generate text and can retrieve enterprise data and trigger actions in business systems with limited human intervention. The survey shows that 14% of organisations have already embedded AI into core operations or autonomous workflows. [10]

An AI agent becomes “rogue” when it acts outside its intended mandate. That may happen because it has been manipulated through prompt injection or poisoned data, because it has been granted excessive permissions, because its credentials have been hijacked, or because it has misread its instructions. Unlike a human insider, an agent acts at machine speed and scale. It may extract and transmit large volumes of personal information, alter or delete records, or spread errors through connected systems before anyone notices. An agent typically operates under its own service identity. Organisations with weak identity controls and incomplete audit trails, which describes the majority of respondents, [11] may be unable to reconstruct what it did.

The technology is novel, but the POPIA analysis is not. Processing by an AI agent that a responsible party deploys is processing by that responsible party, and accountability cannot be delegated to the software. If an agent accesses or discloses personal information outside its authorised purpose, that may constitute unauthorised access triggering section 22 notification obligations. The lack of controls to prevent it may be measured against the section 19 standard.

Agents should therefore be governed as privileged identities:

• give them least-privilege, purpose-bound access;
• log and monitor every action they take;
• require human approval for high-impact actions;
• make sure they can be suspended immediately; and
• back them with tested recovery capability, so that data an agent alters or destroys can be restored to a known clean state.

The market is beginning to recognise this: 41% of organisations are prioritising identity and access controls and 42% AI governance. [12] Readiness of only 29% [13] shows how far there is still to go.

Practical recommendations for clients

The first priority is to stop assuming resilience and start proving it. Organisations should run regular, documented end-to-end recovery tests of their critical data, applications and supporting environments against agreed recovery targets, and fix whatever those tests expose. The evidence those tests produce belongs in the section 19 compliance record, because it is the clearest available proof that safeguards have been verified rather than merely adopted. Recovery testing is only as meaningful as the organisation’s knowledge of what it is recovering. An accurate inventory of the personal and business-critical information held, where it resides and why, should therefore follow. That inventory must extend to backups, SaaS platforms and third parties, and it should be paired with retention schedules that are actually enforced, so that information no longer needed is deleted before it can be stolen.

Knowing where data sits is of limited value without knowing who, or what, has touched it. Clients should be able to produce reliable audit trails of access to their most sensitive datasets across all critical systems. This capability does double duty: it detects misuse, and it makes it possible to establish the facts that a section 22 notification requires. The same discipline must be applied to AI before its use is scaled. That means an acceptable-use policy, a register of approved tools and a personal information impact assessment process for each AI use case. Shadow AI is better addressed through sanctioned alternatives, training and technical controls than through prohibition alone, which tends to drive use out of sight. Where AI systems are given the ability to act, they should be governed as privileged identities. That requires:

• least-privilege, purpose-bound access;
• comprehensive logging;
• human approval for high-impact actions;
• an immediate means of suspension;
• rogue-agent scenarios built into incident response and recovery testing.

These controls lose much of their force if they stop at the organisation’s boundary. Operator agreements should be reviewed to secure section 21-compliant security and breach-notification obligations, meaningful audit rights and a clear allocation of liability. This should be done with particular care for cloud and AI providers and for any processing that engages the cross-border restrictions in section 72. Even well-constructed contracts will not prevent every incident. A tested response plan that brings legal, privacy, communications and technical teams together, with section 22 notification templates and agreed decision-making protocols, is what allows an organisation to meet its obligations under pressure rather than improvise them.

Progress should be measured by recovery test results and by the organisation’s demonstrable ability to explain and control how its data is used. That measurement belongs before the board. Cyber resilience, privacy compliance and AI governance are interdependent board-level risks. Directors are entitled to, and should insist on, regular evidence-based reporting rather than statements of confidence that the survey shows are too often unsupported by results.

Conclusion

The Data Confidence & AI Readiness in Africa Report 2026 is clear: optimism is high, but proven capability lags. For South African organisations, POPIA already provides the framework to close that gap. Its requirements to identify risks, implement and verify safeguards, minimise data, control operators and notify compromises are the building blocks of cyber resilience. AI will be indispensable in meeting those obligations at scale, but only if it is governed with the same discipline. Organisations that run data privacy compliance, resilience and AI governance as one programme will be best placed to withstand the next incident, and to show afterwards that they did.

[1] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[2] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[3] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[4] Data Confidence & AI Readiness in Africa Report 2026, ‘From Confidence to Capability – Veeam Data Platform: Closing the Gaps in Recovery, Visibility, and Governance’, p. 14.

[5] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[6] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans’, pp. 10–11.

[7] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[8] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[9] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[10] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[11] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges’, pp. 7–9.

[12] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans’, pp. 10–11.

[13] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

The post Confidence is not capability (Part 2): Governing AI in the age of the rogue agent appeared first on Werksmans Attorneys.

A two-part series drawing on the Data Confidence & AI Readiness in Africa Report 2026

Introduction

Ransomware is no longer an IT outage. Today’s attacks target backups first, exfiltrate personal information before encrypting it and use the threat of publication to extract payment. Running alongside that threat, artificial intelligence has moved from pilot projects into daily use in most large organisations, usually ahead of the controls meant to govern it. In South Africa, all of this takes place under a data protection regime that has been fully enforceable since 1 July 2021. The Protection of Personal Information Act 4 of 2013 (POPIA) is administered by an Information Regulator that has shown it will act when security safeguards fail.

Boards are now asked a narrower, harder question than whether an incident will occur: can the organisation show, with evidence, that it can withstand one? The Data Confidence & AI Readiness in Africa Report 2026, published by Veeam Software together with ITWeb, suggests that many cannot. Its central finding is that confidence is running well ahead of capability. [1]

This is the first of two articles on the report. Part 1 sets out the key survey findings. It then explains why POPIA compliance operates as a cyber-resilience framework in its own right, and why data visibility is where legal exposure and operational risk meet. Part 2 turns to artificial intelligence, both as a defensive tool and as a source of new risk, including the emerging threat of rogue AI agents, and closes with practical recommendations.

Key findings from the survey

About the survey

The survey was conducted online via the ITWeb and ITWeb Africa platforms during June and July 2026. It drew responses from 140 senior IT, data and cybersecurity leaders at medium and large organisations across 14 African countries, with just under 70% of respondents based in South Africa. [8] Respondents spanned financial services, retail, mining, telecommunications, media and the public sector, and included CIOs, CISOs, CTOs and data and AI leaders. [9] The survey findings are a strong directional indicator, not a statistically definitive picture of the market.

Finding 1: recovery confidence is outpacing recovery proof

More than half (52%) of executives are very or extremely confident that they could recover clean, usable data after a ransomware attack, and a further 33% are moderately confident, so 85% express at least moderate confidence. [10] Yet 32% have never performed a full technical recovery test, have not tested in over a year, or do not know when testing last occurred. [11] Even among organisations that completed an end-to-end recovery test, 45% reported only partial success, significant gaps or outright failure. [12]

The report concludes that some organisations are relying on backup policies or assumptions rather than validated recovery performance. It warns that, without regular end-to-end testing, a ransomware incident could result in extended downtime, data-integrity problems, regulatory exposure and reputational damage. [13]

In legal terms, this is the difference between having a safeguard on paper and being able to demonstrate that it works.

Finding 2: AI adoption is outpacing organisational controls

Only 29% of organisations consider themselves advanced or fully ready to adopt, govern and scale AI using trusted data. [14] At the same time, 36% use approved AI tools in selected business functions, 27% are running pilots or experiments, and 14% have embedded AI into core business operations or autonomous workflows. [15] A further 11% report informal or employee-led use of public AI tools with limited oversight. [16]

Around half of respondents use AI in business functions, core processes, or automated workflows, yet fewer than a third consider themselves ready to govern it. The report treats the presence of “shadow AI” as evidence that employee demand is moving faster than organisational controls. [17]

Without stronger data governance, access controls and oversight, organisations risk exposing sensitive information, creating compliance issues and undermining trust in AI-generated outputs. [18]

Finding 3: organisations cannot fully see or account for their data

The biggest barriers to maintaining data confidence are skills shortages (32%), incomplete or inconsistent data classification (28%), budget constraints (27%), lack of visibility into sensitive or critical data (26%) and legacy systems (25%). [19] Respondents also cite data sprawl across cloud, SaaS, endpoints, backups and third parties (23%), reliance on manual inventories or spreadsheets (19%) and shadow IT or shadow AI use (17%). [20]

The accountability gap is stark. Only 37% say they can produce an accurate audit trail showing who accessed their most sensitive datasets across most or all critical systems in the past seven days; 55% can do so only partially or for selected systems, and 8% cannot or do not know. [21] The report observes that fragmented visibility makes it difficult to demonstrate compliance, detect misuse or respond quickly to security incidents. It adds that organisations without accurate data inventories and access monitoring will struggle as regulatory scrutiny intensifies. [22]

Finding 4: investment is accelerating in the right places

Nearly 70% of organisations expect to increase spending on data protection, cyber resilience and AI governance over the next 12 months, with 30% anticipating a significant increase. [23] Investment priorities are backup and recovery (54%), AI governance (42%), identity and access controls (41%), data discovery and classification (36%) and cloud and SaaS data protection (30%). [24] Privacy automation (15%) and data security posture management (6%) remain lower priorities. [25] The report notes that organisations are beginning to treat cyber resilience and AI as two outcomes of the same foundation, trusted data, rather than as separate initiatives. [26]

Finding 5: geography and size shape maturity, but not always as expected

Multi-country organisations are markedly more mature. They are more likely to have AI embedded in core operations (22% against 5%) and to be advanced or fully AI-ready (39% against 18%); they report higher ransomware recovery confidence (60% against 45%) and fewer unsuccessful recovery tests (4% against 11%). [27] The report attributes this to the greater regulatory and operational complexity of cross-border operations, which drives earlier investment in AI governance, data visibility and cyber resilience, although multi-country organisations struggle more with data sprawl, classification and cross-border compliance. [28] Single-country organisations are more constrained by skills shortages (40% against 24%) and budgets, and focus on backup and recovery (67% against 42%). [29]

Size, by contrast, does not guarantee maturity. Smaller organisations (under 500 employees) are more likely to be very confident that they know where sensitive data resides (47% against 30%) and rate their data-management capability more positively (56% against 38%); no smaller respondent rated its capabilities as poor, compared with 11% of large organisations. [30] Larger organisations place greater emphasis on identity and access controls (45% against 33%), while smaller organisations invest more heavily in AI governance (51% against 38%) and skills development (37% against 17%). [31]

The lesson for South African clients is that regulatory complexity, properly managed, appears to drive maturity. Organisations that have seriously dealt with cross-border data protection obligations are better prepared to recover from attack and scale AI safely.

The imperative of data privacy compliance: POPIA as a cyber-defence framework

Privacy compliance and cybersecurity are still often treated as separate disciplines, one owned by legal and the other by IT. The survey shows why that division no longer holds. Almost every weakness it identifies corresponds directly to an existing obligation under POPIA: untested recovery, poor data visibility, inadequate audit trails, uncontrolled AI use and data sprawl across third parties. Properly implemented, compliance is not an overhead added to security. It is a structured security programme that the law already requires.

Section 19: security safeguards are a continuous, risk-based obligation

Section 19 of POPIA requires a responsible party to secure the integrity and confidentiality of personal information in its possession or under its control. It must do so by taking appropriate, reasonable technical and organisational measures to prevent loss of, damage to or unauthorised destruction of personal information, and unlawful access to or processing of it. The section also requires the responsible party to identify all reasonably foreseeable internal and external risks, to establish and maintain appropriate safeguards against them, to regularly verify that those safeguards are effectively implemented, and to ensure that they are continually updated in response to new risks or deficiencies. It also requires due regard to generally accepted information security practices.

The duty to verify is where the law and the survey meet. An organisation that has never conducted an end-to-end recovery test, or whose last test failed, will find it difficult to show that it has regularly verified the effectiveness of its safeguards. That applies to the 32% of respondents who have not recently tested, or do not know when they last did, [32] and to the 45% of testers who reported partial success, significant gaps or failure. [33] These organisations are not only operationally exposed; they may also face a finding that they have not met the section 19 standard. Ransomware that destroys or encrypts personal information without a reliable route to restoration is exactly the “loss” and “damage” the section is designed to prevent.

Know your data: the processing conditions shrink the attack surface

Several of POPIA’s conditions for lawful processing operate, in practice, as attack-surface reduction measures. The minimality principle (section 10) limits processing to what is adequate, relevant and not excessive. The retention and restriction provisions (section 14) require records to be kept no longer than necessary and then destroyed, deleted or de-identified. Information that has been lawfully deleted cannot be stolen, encrypted or published. Organisations struggling with data sprawl across cloud, SaaS, endpoints, backups and third parties (23%) and with inconsistent classification (28%) [34] are carrying breach liability they do not need to carry.

Compliance with POPIA also presupposes that an organisation knows what personal information it holds, where it is and why it is processed. Two sets of requirements force exactly the data mapping and classification that the survey identifies as weak. The first is the Information Officer’s duties under section 55. The second is regulation 4 of the POPIA Regulations, which requires a compliance framework, personal information impact assessments and related measures. The report’s own recommendation is to know what data is held, where it sits, how it is used and which privacy requirements apply, including where it moves across borders. [35] In substance, that describes a mature POPIA compliance programme.

Section 22: you cannot notify what you cannot see

Section 22 applies where there are reasonable grounds to believe that personal information has been accessed or acquired by an unauthorised person. The responsible party must then notify the Information Regulator and, subject to limited exceptions, the affected data subjects as soon as reasonably possible after discovery. The notification must give data subjects enough information to take protective measures, including a description of the possible consequences, the measures taken or intended, and, where known, the identity of the unauthorised person.

That obligation cannot be met without evidence. Most respondents cannot produce an accurate audit trail showing who accessed their most sensitive datasets across most or all critical systems. That is the position of 63%: 55% can do so only partially or for selected systems, and 8% cannot or do not know. [36] Organisations in that position will struggle to determine whether a compromise has occurred, how far it extends and whom to notify. The result is either delayed notification or over-notification, and both carry regulatory and reputational cost. Access logging and monitoring are therefore more than security controls; they are the evidence on which lawful incident response depends.

Operators, third parties and cross-border flows

Sections 20 and 21 govern operators, the service providers that process personal information on a responsible party’s behalf. An operator may process the information only with the responsible party’s knowledge or authorisation and must treat it as confidential. It must also be bound by a written contract requiring it to establish and maintain section 19 security measures and to notify the responsible party immediately of any suspected compromise. Cloud, SaaS and outsourced services are now the norm, and the survey finds that data sprawl extends to third parties. [37] Supply-chain contracting and oversight are therefore among the most effective cyber-risk controls available. Section 72 restricts transfers of personal information outside South Africa unless prescribed safeguards are met. It applies to multi-country organisations, which report greater difficulty with cross-border compliance, [38] and to employees who paste personal information into public AI tools hosted offshore.

Enforcement and accountability raise the stakes

POPIA’s accountability condition (section 8) places responsibility squarely on the responsible party. The Information Regulator may issue enforcement notices and infringement notices, and administrative fines of up to R10 million may be imposed. The Regulator has already taken enforcement action against a national government department following a ransomware attack, citing failures in basic security safeguards. Sector-specific regimes add further obligations; financial institutions, for example, are subject to the Prudential Authority and FSCA Joint Standard on cybersecurity and cyber resilience. Across these regimes, regulators expect evidence that safeguards work, not assurances that they exist.

POPIA compliance reduces exposure to data breaches, ransomware and other cyber threats in four concrete ways. It forces organisations to identify and minimise the data at risk. It requires risk-based safeguards that are regularly verified. It extends security obligations contractually into the supply chain. It also builds the visibility and governance needed to detect, contain and report incidents quickly.

Part 1 in summary

The survey describes a market that is confident but not yet able to prove that confidence. Recovery is assumed more often than it is tested, and when it is tested, it often fails. [39] Fewer than four in ten organisations can account for who has accessed their most sensitive data across most or all critical systems. [40] Investment is rising, and it is increasingly directed at governance and visibility rather than backup alone.

POPIA does not create these gaps, but it turns them into legal exposure. Sections 19 and 22, the processing conditions and the operator regime already require what resilient organisations do anyway: they know their data, verify their safeguards, control their suppliers and can explain an incident when it occurs.

Data visibility is the common thread. It determines whether an organisation can minimise what it holds, verify what it protects and notify what it loses. It will also determine whether an organisation can safely deploy the AI tools now entering its systems.

In Part 2, we examine the role of artificial intelligence in cybersecurity. We consider how it can strengthen detection, recovery and compliance, how shadow AI creates new exposure under POPIA, and why autonomous AI agents acting outside their mandate are the next governance problem for boards. We close with practical recommendations for clients.

[1] Data Confidence & AI Readiness in Africa Report 2026, ‘From Confidence to Capability’, p. 14.

[2] Data Confidence & AI Readiness in Africa Report 2026, ‘Overview of Survey Methodology’, p. 2.

[3] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability’, p. 3.

[4] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[5] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[6] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges’, pp. 7–9.

[7] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans’, pp. 10–11.

[8] ‘Overview of Survey Methodology’, p. 2.

[9] ‘Overview of Survey Methodology – Respondent Profile’, p. 2.

[10] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability’, p. 3.

[11] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability’, p. 3.

[12] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[13] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[14] Data Confidence & AI Readiness in Africa Report 2026 ‘AI Adoption Outpaces Readiness’, p. 5.

[15] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[16] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness’, p. 5.

[17] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness – Employee Use of AI is Outpacing Organisational Controls’, pp. 5–6.

[18] Data Confidence & AI Readiness in Africa Report 2026, ‘AI Adoption Outpaces Readiness – Employee Use of AI is Outpacing Organisational Controls’, pp. 5–6.

[19] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[20] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[21] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges’, pp. 7–9.

[22] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Limited Visibility and Access Monitoring Undermine Data Protection, Governance and Trust’, p. 9.

[23] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans’, pp. 10–11.

[24] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans’, pp. 10–11.

[25] Data Confidence & AI Readiness in Africa Report 2026,  ‘Investment Plans’, pp. 10–11.

[26] Data Confidence & AI Readiness in Africa Report 2026, ‘Investment Plans – Spending Plans Suggest that Trusted Data is Becoming a Strategic Business Priority’, p. 11.

[27] Data Confidence & AI Readiness in Africa Report 2026, ‘Geography Shapes AI and Resilience Maturity’, p. 12.

[28] Data Confidence & AI Readiness in Africa Report 2026, ‘Geography Shapes AI and Resilience Maturity’, p. 12.

[29] Data Confidence & AI Readiness in Africa Report 2026, ‘Geography Shapes AI and Resilience Maturity’, p. 12.

[30] Data Confidence & AI Readiness in Africa Report 2026, ‘Organisational Size Does Not Equal Maturity’, p. 13.

[31] Data Confidence & AI Readiness in Africa Report 2026, ‘Organisational Size Does Not Equal Maturity’, p. 13.

[32] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability’, p. 3.

[33] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[34] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[35] Data Confidence & AI Readiness in Africa Report 2026, ‘From Confidence to Capability’, p. 14.

[36] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges’, pp. 7–9.

[37] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges – Biggest Challenges in Maintaining Confidence in Data’, pp. 7–8.

[38] Data Confidence & AI Readiness in Africa Report 2026, ‘Geography Shapes AI and Resilience Maturity’, p. 12.

[39] Data Confidence & AI Readiness in Africa Report 2026, ‘Confidence vs. Capability – Recovery Confidence is Outpacing Recovery Proof’, pp. 3–4.

[40] Data Confidence & AI Readiness in Africa Report 2026, ‘Data Visibility and Audit Challenges’, pp. 7–9.

The post Confidence is not capability (Part 1): What the data shows, and why POPIA is the foundation of cyber resilience appeared first on Werksmans Attorneys.

The South African Bureau of Standards (SABS) has published a series of draft standards on artificial intelligence (AI) for public comment.

The drafts adopt international ISO/IEC standards as South African documents, and cover transparency, data quality, human oversight, and the benefits and environmental impact of AI. While the standards are voluntary, they are likely to become the local benchmark for responsible AI, in line with international best practice, creating alignment between South Africa and global standards.

Interested parties, including organisations that develop or use AI, may submit comments by 10 November 2026.

The SABS has published draft standards, technical specifications and technical reports governing the use of AI in South Africa.

The drafts were prepared by the SABS Technical Committee on Information Technology and Artificial Intelligence (SABS/TC 001/SC 42), which has proposed that the documents should become South African standards, technical reports and technical specifications.

Each draft is an identical adoption of an existing ISO/IEC document on artificial intelligence, without local changes. In other words, the international text is replicated in the South African documents.

The International Organization for Standardization (ISO) and the International Electrotechnical Commission (IEC) are independent international bodies that develop standards through experts nominated by ISO member bodies. Their joint AI committee is ISO/IEC JTC 1/SC 42, which the local SABS committee mirrors. Because ISO/IEC standards are developed by consensus across many countries, they are widely recognised by regulators, investors and business partners, and are often used as a reference point for good practice.

The identical adoption of ISO/IEC standards is a worldwide norm. It promotes conformity with international benchmarks used by trading partners, and enables smaller countries to benefit from the work done by experts from many countries, rather than each country writing its own standards, promoting international consistency and certainty.

The drafts seek to create voluntary standards and specifications for organisations deploying AI, in the following four areas:

• Transparency (SANS 12792 / ISO/IEC 12792:2025): This sets out a common structure for the information that should be disclosed about an AI system, covering the system itself, the underlying model, the datasets used and the wider societal and environmental context. The aim is to help developers, deployers, auditors, authorities and end users understand how an AI system works, what it is meant to do and what its limits are.
• Data quality (SANS 2620 Parts 1, 2, 3 and 5 / ISO/IEC 5259-1, -2, -3 and -5): Data quality is critical to reliable AI, because AI systems learn from data sets. These parts set out key concepts and the data life cycle, measures for data quality characteristics such as accuracy, completeness and representativeness, requirements for managing data quality across the life cycle, and a governance framework for boards and management to direct and oversee data quality.
• Human oversight (SATS 8200 / ISO/IEC TS 8200:2024): This deals with the controllability of automated AI systems. It sets out a framework so that a human or other controller can monitor the state of an AI system and take over control when needed, with different levels of controllability and steps to test that these controls work.
• Benefit and environmental impact (SATR 24030 / ISO/IEC TR 24030:2024 and SATR 20226 / ISO/IEC TR 20226:2025):
  • The use case report collects 81 real-world AI use case summaries across sectors such as fintech, e-commerce, energy, agriculture, education, healthcare, insurance, manufacturing, media and entertainment, security and transport.
  • The environmental report gives an overview of the environmental footprint of AI systems across their life cycle and supply chain, including energy use, carbon, water and e-waste, and maps possible metrics and actions to reduce that impact.

Such standards are not legally binding (unless they have been made compulsory, for example by being referenced in law). Rather, they create voluntary standards which organisations can choose to adopt to improve quality, safety, transparency and the like, in accordance with the standards developed by internationally recognised experts.

How to comment

Anyone can comment on the draft standards by 10 November 2026, by email to dsscomments@sabs.co.za.

Comments must be submitted using the SABS commenting template (SABS_commenting_template.doc), with reference to the SABS User Guide.

The post SABS publishes draft AI standards for public comment appeared first on Werksmans Attorneys.

The Labour Court’s decision in Bosch v Passenger Rail Agency of SA provides employers with a useful warning: when designing voluntary severance or early-retirement programmes, age may be an administratively convenient dividing line, but convenience does not make discrimination lawful. The dispute arose from PRASA’s efforts to reduce its wage bill and headcount. It introduced a voluntary severance package (“VSP”) for employees generally, while employees between the ages of 60 and 63 were instead offered voluntary early retirement (“VER”) where the VSP would have produced a higher amount. The financial distinction was significant. Employees younger than 60 could receive a VSP incorporating a “sweetener” of 2.5 weeks’ salary for every completed year of service. Employees between 60 and 63 were limited to a lump sum calculated according to the months remaining until retirement.

Bosch was 62 and had more than 44 years’ service. He initially obtained a VSP quote exceeding R2 million and submitted his application. He was subsequently informed that his application had been treated as one for early retirement and that he would receive R657,906.67 – effectively seven months’ salary plus notice pay.  After unsuccessfully pursuing PRASA’s internal grievance process, Bosch challenged the distinction as unfair discrimination based on age under section 6(1) of the Employment Equity Act (“EEA”). The difficulty for PRASA was straightforward: the different treatment was expressly determined by age. The Court found that the only reason Bosch received the substantially lower VER package rather than the VSP was that he was over 60.

Once discrimination on a listed ground such as age is established, section 11(1) of the EEA places the burden on the employer to prove that the discrimination did not occur, or that it was rational and not unfair, or otherwise justifiable. PRASA struggled at this crucial stage. Its witnesses had not made the decision to differentiate between employees according to age and could not explain why the board had done so. No board member, or anyone else involved in the decision, testified. The Court was therefore left without admissible evidence explaining why the distinction had been adopted or demonstrating the savings supposedly achieved by it.

PRASA also argued that Bosch could have withdrawn his application once he knew what he would receive and that, by accepting the VER payment, he had effectively waived his right to complain. The Court rejected this argument. It held that Bosch’s acceptance of the payment did not waive his right to challenge the discriminatory treatment. PRASA’s reliance on a “full and final settlement” provision similarly could not rescue the differentiation. This is an important practical point for employers. A signed separation document should not be treated as an automatic cure for an underlying discrimination problem.

The Court ultimately held that PRASA’s policy breached the EEA and that the failure to pay Bosch the VSP constituted unfair age discrimination. PRASA was ordered to pay a global amount of approximately R1.59 million, subject to tax, together with interest and costs. The judgment does not mean that employers can never differentiate between employees when designing restructuring, severance or retirement arrangements. It does, however, emphasise that where the differentiation is directly linked to a listed ground such as age, the employer must be able to justify both the distinction and the rationale underlying it with evidence. For employers contemplating voluntary retrenchment or early-retirement programmes, that justification should therefore be interrogated – and documented – when the scheme is designed, rather than reconstructed when the matter reaches court.

The post Too old for the package? Labour Court draws the line on age-based severance appeared first on Werksmans Attorneys.

Introduction

The Gauteng Division of the High Court of South Africa recently published a notice of proposed amendments (“Proposed Amendments“) to its Mandatory Mediation Protocol (“Mediation Protocol”), which operationalises the Division’s Mandatory Mediation Directive. The proposed amendments are an attempt to reduce the clogging in the Special Interlocutory Court (“SIC“).

The Purpose of the SIC

The SIC was established to address procedural non-compliance by an adversary in litigation. The Division’s practice directive states that “the very purpose of the SIC is to overcome delinquency and consequent improper delays”. It further provides that any litigant who is aggrieved by another litigant’s “neglect, dilatoriness, failure, or refusal to comply” with any provision of the Uniform Rules of Court (“Rules“), practice manual, or the practice directive must approach the SIC to compel compliance from the delinquent party.

Notably, the orders obtainable in the SIC are strictly of an interlocutory nature and therefore do not deal with the substantive (or main) aspects of the litigation. By having a dedicated court of this nature, it was envisaged that there would be a reduction of the time spent on interlocutory issues which would create more capacity for other court rolls to deal with substantive issues.

Unfortunately, despite best intentions, the SIC as an innovation has faced a significant functional problem – excessive congestion. The Proposed Amendments highlight that at any given time there are approximately 16 000 matters seeking allocation and hearing dates in the SIC, which causes congestion, delays and backlogs.

In some ways, this congestion is a self-fulfilling prophecy. Mandatory Mediation, pursuant to the procedures in the Mediation Protocol, was meant to free up the court rolls and foster the expedient and negotiated settlement of cases between parties. Instead, the Mediation Protocol simply moved the traffic from one intersection to another, and the SIC suffers for it. Resultantly, despite the relative availability of court dates, the cases that previously clogged up the court rolls did not disappear, they have now just clogged up the system meant to expediently resolve procedural issues. Accordingly, the proposed amendments are necessitated by the SIC backlog of procedural delinquency cases, owing to parties not fully complying with the Mediation Protocol.

The Mediation Protocol

The introduction of mandatory mediation in Gauteng, accompanied by the Mediation Protocol to operationalise the system, also made peremptory the rules of procedure pertaining to mediation.

From the perspective of the Gauteng Division, mediation was meant to promote alternative dispute resolution and ensure that only those cases which absolutely require judicial intervention come before the judges of the Division. However, alternative dispute resolution processes require the active participation of all parties to a dispute, failing which the peremptory mediation rules are undermined. Accordingly, and as an example, where a party fails to respond to a notice calling for mediation, that constitutes non-compliance and such delinquency may be brought before the SIC for what should ideally be expedient resolution. The Proposed Amendments explain that the SIC congestion is primarily due to litigants’ non-responsiveness to Rule 41A mediation notices. This not only undermines the underlying rationale of the SIC which, in addition to what is stated above, includes the prevention of delay and gamesmanship but also the effective implementation of the Mediation Protocol.

In Malebane v Road Accident Fund [2025] ZAGPPHC 1253, the court remarked that of approximately 200 applications serving before the SIC during that week, a significant number were “Delinquency Applications”. The court further noted that most, if not all, of these applications arose from respondents’ failures to respond to Rule 41A mediation notices. The court expressed growing concern that delinquency applications were increasingly being used as tactical shortcuts to gain access to the Default Judgment Roll. This is an example of the exact gamesmanship the SIC is meant to prevent. The court’s observations are particularly relevant to the proposed amendments. The fact that the majority of matters before the SIC comprise of applications of this nature underscores the need for measures aimed at addressing their increase and lessening the misuse of the delinquency mechanism. The proposed amendments are an attempt to address this issue.

The Proposed Amendments

The solution suggested in the Proposed Amendments is to excise matters where the only issue is a litigant’s failure to reply to a Rule 41A mediation notice. In terms of the Proposed Amendments, mediation delinquency would instead be dealt with through the procedure prescribed in Rule 30A of the Rules which governs non‑compliance with the Rules and Court Orders. This marks a departure from the existing procedure in the Mediation Protocol, which requires an aggrieved party to obtain an order from the SIC compelling the delinquent party to comply with the Mediation Protocol, failing which the aggrieved party may seek that: the delinquent litigant’s claim or defence be struck out, the matter be enrolled on the Default Judgment Roll, and the delinquent litigant be held to be in contempt of the compelling order.

In terms of the proposed solution where a party fails to deliver a response to a Rule 41A mediation notice, the aggrieved party (who delivered such notice) may serve a Rule 30A Notice calling upon the delinquent party to respond within 10 days of delivery of that notice, failing which the aggrieved party may apply to the relevant Registrar for the allocation of the matter on the Default Judgment Roll or the Special Default Judgment Roll if the delinquent party fails to deliver a response after 15 days of expiry of the period called for in the Rule 30A Notice.

Conclusion

While the proposed amendments may result in a reduction of the congestion in the SIC and support the effective implementation of the mandatory mediation regime, they may also just result in another unsuccessful re-direction of judicial traffic; this time from the SIC to the Default Judgment Roll and the Special Default Judgment Roll. It is contemplated in the Proposed Amendments that if the proposed solutions are approved, then transitional provisions to facilitate their pragmatic implementation should be developed and adopted.

It is hoped that the proposed changes have the desired effect and improve the administration of justice in the Gauteng Division. The importance of effective judicial administration cannot be understated, especially given the Gauteng Division’s place as the busiest court in the country.

The post Proposed Amendments to the Mandatory Mediation Protocol in the Gauteng Division of the High Court: Will the latest attempt to redirect judicial traffic succeed? appeared first on Werksmans Attorneys.

1. Section 11 of the Mineral and Petroleum Resources Development Act 28 of 2002 (“the MPRDA”) is one of the most frequently invoked provisions in the mining regulatory landscape. Every cession, transfer or disposal of a prospecting or mining right, or of an interest in such a right is subject to its provisions. It is, however, easy to lose sight of what the consent process is actually designed to do. A recent judgment of the High Court of South Africa, Gauteng Division, Pretoria, serves as a timely reminder that ministerial consent is a regulatory approval of a transaction, not a cure for a transaction that never lawfully came into existence.

2. In Misty Falls 45 (Pty) Ltd and Another v The Minister of Mineral Resources and Energy and Others (Case No. 19058/2022) (“the Misty Falls Judgment”), handed down by Raubenheimer AJ and deemed delivered on 10 April 2026, the Court reviewed and set aside the Director-General’s (“DG”) consent to the transfer of the Kareepan mining right, as well as the Minister’s dismissal of the internal appeal against that consent. Unusually, the Court went further and substituted its own decision: consent was refused.

3. By way of background, Misty Falls 45 (Pty) Ltd (“Misty Falls 45”) holds a manganese and iron ore mining right over portions of the farms Kareepan 450 and Pensfontein 449 in the Kuruman district. A former director and shareholder (“the former director”) sought, together with a director of Misty Falls 111 (Pty) Ltd (“Misty Falls 111”), consent under section 11 to transfer that right to Misty Falls 111.

4. The former director relied on a 2010 “Exchange of Rights Agreement” between shareholders and a 2011 shareholders’ resolution as the source of his authority. The documentary record, however, showed that he had resigned as director in 2014. No special resolution in terms of sections 112 and 115 of the Companies Act 71 of 2008 was ever passed, and the then board expressly told the Department that no transaction involving the right should be processed without its permission.

5. The underlying causa also shifted over time. A February 2018 letter referred to a “cession agreement” between Misty Falls 45 and Misty Falls 111 which did not exist. A 2019 motivation letter relied instead on a November 2018 agreement between the former director personally and Misty Falls 111.

6. The DG nonetheless granted consent in January 2020, focusing on the transferee’s financial and technical capacity without engaging with the questions of authority and causa. On appeal, the Department’s Chief Director: Legal Services identified authority as the “crisp issue” and recommended that the appeal be upheld. The Minister dismissed the appeal two days later without engaging with that recommendation.

7. Section 11(1) provides that a prospecting or mining right, or an interest in it, may not be ceded, transferred, let, sub-let, assigned, alienated or otherwise disposed of without the Minister’s written consent. Section 11(2) provides that consent “must be granted” if the transferee can comply with the obligations and conditions of the right and satisfies the requirements of section 17 or section 23, as the case may be. On its face, therefore, the enquiry is directed at the transferee and its capacity, and in practice that is where the attention of the Department is usually focused.

8. The Court, however, identified a logically prior question. As Raubenheimer AJ put it, section 11 “presupposes the existence of a lawful cession, transfer, alienation or disposal of a right”, and the Minister’s consent “cannot validate an otherwise unlawful or unauthorised transaction”. Put simply, there must first be a valid transaction to which consent can attach.

9. Applying that principle, the Court found that none of the sources of authority relied upon could sustain the transfer-

9.1. The Exchange of Rights Agreement dealt only with the allocation of economic benefits from two prospecting rights among shareholders. Section 19(1)(b) confers on the holder of the prospecting right the exclusive right to apply for a mining right – this right is not conferred on individual shareholders. Reading the agreement as transferring a future mining right would render it unlawful as an attempt to circumvent section 11.

9.2. The 2011 resolution, properly construed, was aimed at establishing and operating the mine, not at disposing of the right itself. In any event, a private resolution cannot override the peremptory requirements of sections 112 and 115 of the Companies Act.

9.3. Whatever authority the former director held as a director lapsed upon his resignation, and a shareholder, even a majority shareholder, has no inherent authority to bind a company to the disposal of its assets.

10. The review Court accordingly concluded that the section 11 application was neither authorised by Misty Falls 45 nor supported by a valid underlying transaction, and that the consent was not authorised by the empowering provision. In considering remedy, the Court described the “central defect” as “the absence of both authority and a valid underlying transaction” and held that there was “no conceivable basis” on which the DG or the Minister could lawfully grant consent on the facts. Refusal was the only lawful outcome, and remittal “would serve no purpose other than to delay the inevitable”.

11. The Court also addressed the limits of the regulator’s role. The Minister is not empowered to adjudicate private disputes about shareholding, directorship or internal corporate arrangements, save to the extent necessary to determine whether the application is duly authorised and lawful. Faced with manifest disputes about authority, the Minister was required either to refuse consent or to insist that the parties resolve their dispute in the appropriate forum. By “siding” with one faction on contested questions of corporate law, the Minister exceeded the scope of his section 11 powers. In the Court’s words, the regulator’s role “is not to confer legitimacy on contested private arrangements that may themselves be unlawful or ultra vires”.

12. The significance of the Misty Falls Judgment lies in its clear statement that a void transaction cannot be rescued through section 11. Consent is not a form of administrative ratification. The Court was careful to note that nothing prevents the parties from structuring a lawful, properly authorised transaction in future and applying afresh, but what is impermissible is “to validate, through administrative fiat, an unauthorised and opaque transaction”. The Court also held that, while the capacity assessment under section 23(1)(b) attracts a more deferential standard, decision-makers remain obliged to consider whether the underlying transaction and authority are sound.

13. For the industry, the implications are practical. Regulators processing section 11 applications cannot confine themselves to a box-ticking assessment of the transferee’s financial and technical capacity, particularly where objections from the right holder’s board are on record. Practitioners should ensure that applications are supported by a written transaction concluded by the right holder itself, proper board authority and, where the right constitutes all or the greater part of the company’s assets, the requisite special resolutions. Market participants acquiring rights should treat the existence of internal disputes as a serious red flag, since a consent obtained on a defective foundation is vulnerable to being set aside, with adverse costs to follow.

14. Whilst the Misty Falls Judgment is binding only in the Gauteng Division, it is persuasive elsewhere and restates a principle of general application: section 11 regulates lawful transactions, it does not create them. Where the foundation is absent, there is simply nothing for the regulator to consent to.

The post Section 11 consent: Can the regulator breathe life into a void transaction? appeared first on Werksmans Attorneys.

Constructive dismissal is not an escape route for employees who are dissatisfied with their working environment, nor does every workplace grievance entitle an employee to resign and demand compensation. The Labour Appeal Court’s recent judgment in Royale Energy (Pty) Ltd v Commission for Conciliation, Mediation & Arbitration & others (2026) 47 ILJ 2419 (LAC) provides a welcome reminder that intolerability is a high threshold, and that employers must be afforded a reasonable opportunity to address workplace complaints before employees resort to resignation.

The employee, a senior manager at Royale Energy, had experienced a difficult relationship with the company’s group chief executive officer over several years. Following an incident in August 2020, during which the CEO allegedly behaved aggressively and threateningly towards her, she lodged a formal grievance and requested the board’s intervention.  The employer responded by appointing an external investigator and proposing alternative reporting arrangements while the investigation continued. The employee, however, rejected the proposed arrangements, sought a mutual separation package exceeding R4 million and, when this was declined, demanded the investigation report.

Less than a month after lodging her grievance, and before the investigation had been completed, she resigned and referred a constructive dismissal dispute to the CCMA.  The commissioner found in her favour, a decision subsequently upheld by the Labour Court, which awarded compensation of approximately R1.15 million.

The Labour Appeal Court took a very different view and held that intolerability is not simply unhappiness. Section 186(1)(e) of the Labour Relations Act provides that constructive dismissal occurs where an employee terminates employment because the employer has made continued employment intolerable.  The Court reaffirmed that this requires an objective assessment. It is insufficient for an employee merely to believe that the workplace has become unbearable. The employer’s conduct must objectively render continued employment intolerable.

Importantly, the Court distinguished between an unpleasant or difficult working relationship and circumstances genuinely leaving an employee with no reasonable alternative but to resign. The employee had continued working despite several earlier incidents involving the CEO, none of which had been formally escalated to the board. When she eventually complained, the employer took steps to address her concerns. These facts were inconsistent with her contention that continued employment had become objectively intolerable.

Perhaps the most significant aspect of the judgment is the Court’s treatment of the employer’s response to the grievance. The employee had expressly requested intervention and indicated her willingness to participate in a process aimed at resolving the dispute. Yet, when the employer initiated precisely such a process, she resigned before allowing it to run its course.  The Court was particularly critical of the employee’s attempt to use the ongoing remedial process itself as evidence of intolerability. An employer cannot reasonably be expected to investigate allegations spanning several years, consider the competing versions and implement appropriate corrective measures within an arbitrarily short period.  Nor can an employee invoke a grievance procedure, refuse to allow it to reach its conclusion, and then rely on the absence of an immediate resolution as justification for resignation.

The Court also rejected the employee’s reliance on anticipated future misconduct. Constructive dismissal cannot be founded on speculation that working conditions might become intolerable at some future point – the employer must be given a fair opportunity to resolve the complaint.  Employers are entitled to investigate complaints, implement interim protective measures and explore workable solutions without automatically exposing themselves to constructive dismissal claims. This does not excuse genuine workplace bullying or permit employers to ignore legitimate grievances. It does, however, recognise that a properly constituted remedial process requires time and cooperation. The Labour Appeal Court found that no constructive dismissal had occurred.

For employers, the message is clear. A prompt, documented and reasonable response to workplace grievances can provide a powerful defence against constructive dismissal claims.  For employees, the lesson is equally important: a resignation does not become a dismissal merely because the employee is unhappy with the employer’s response.  The law protects employees against genuinely intolerable working conditions. It does not guarantee an immediate or preferred outcome to every workplace complaint.

The post Resign First, Complain Later? The LAC Draws the Line on Constructive Dismissal appeared first on Werksmans Attorneys.

Environmental authorisations impose strict obligations which continue throughout the life of a project. In instances where unlawful pollution persists, preparing remedial plans does not discharge those obligations, and regulators cannot remain inactive in the face of non-compliance. In Topigs Norsvin (Pty) Ltd v Eskom Holdings SOC Ltd and Others, the Supreme Court of Appeal (“SCA“) considered whether pollution downstream of the Kusile Power Station (“Kusile“) justified court orders compelling Eskom Holdings SOC Ltd (“Eskom“) and the responsible authorities to comply with their environmental duties.

Factual Background

Topigs Norsvin SA (Pty) Ltd (“Topigs“) operates a swine genetics business on a farm adjacent to the Kusile Power Station in Mpumalanga. Eskom is responsible for Kusile’s construction and operation. Kusile’s activities were subject to environmental authorisations, an approved environmental management programme and water-use licences, which imposed requirements governing pollution prevention, water quality monitoring and the management of contaminated water.

Through its representative on Kusile’s Environmental Monitoring Committee, Topigs repeatedly raised concerns about inadequate monitoring and non-compliance. Reports were delayed or deficient and undertakings to address pollution were repeatedly left unfulfilled. Eskom acknowledged that its operations had adversely affected downstream water resources.

Topigs approached the Gauteng Division of the High Court, Pretoria (“High Court”) for declaratory and mandatory relief, together with a structural interdict requiring continuing judicial supervision and reports every 15 days. The High Court dismissed the application, finding that section 28(12) of the National Environmental Management Act 107 of 1998 (“NEMA“) provided an alternative remedy which Topigs had not pursued, and that its reliance on constitutional rights offended the principle of subsidiarity. Topigs then appealed to the SCA.

SCA Appeal

The SCA rejected the contention that Topigs had improperly bypassed environmental legislation by relying on section 24 of the Constitution. The principle of subsidiarity generally requires a litigant to rely on legislation enacted to give effect to a constitutional right. However, Topigs’ case was substantively founded on NEMA, the National Water Act 36 of 1998 (“NWA”) and breaches of Eskom’s authorisations and licences. Furthermore, section 28(12) of NEMA was not an exclusive remedy. Section 32(1) permits an aggrieved person to seek appropriate relief for breaches or threatened breaches of environmental legislation. The failure expressly to invoke section 28(12) did not defeat proceedings firmly grounded in that statutory framework. The SCA found that Topigs had established a clear right, actual or reasonably apprehended harm, and the absence of another satisfactory remedy.

On the merits, Eskom’s persistent non-compliance and the authorities’ inadequate enforcement established breaches of statutory duties and section 24 of the Constitution. The SCA held that the enforcement powers in section 28(4) of NEMA could not be treated as discretionary in the face of established non-compliance. The responsible officials were required to act.

The SCA furthermore rejected the suggestion that this was merely a dispute between neighbours. The protection of water resources and the lawful exercise of public power raised matters of public interest, with pollution capable of affecting communities beyond the adjoining properties.

However, the SCA distinguished between compelling compliance and assuming ongoing supervision of technical remedial work. Updated action plans and steps taken by the departments after litigation commenced indicated a willingness to address their responsibilities. The structural interdict sought by Topigs was therefore inappropriate in the circumstances.

This did not justify refusing all relief. The SCA upheld the appeal in part and granted declarations of unlawfulness and constitutional breach, together with mandatory orders and reporting obligations. Eskom and its current board were directed to submit a detailed action plan under oath within 30 days of the order, identifying the necessary steps, responsible officials and timeframes. They were required to implement the plan without delay and report under oath within 60 days of the order on compliance, explaining any failures and proposed remedial steps.

The Directors-General of the Department of Water and Sanitation and the Department of Forestry, Fisheries and the Environment were ordered to take steps to ensure compliance. Furthermore, within 75 days of the order, they had to file affidavits confirming that they had considered and determined whether enforcement directives, compliance notices or remedial measures should be issued against Eskom under NEMA and the NWA. Topigs was granted leave to respond to Eskom’s reports and approach the court for further compliance relief.

Topigs did not, however, succeed against every respondent. The SCA upheld the misjoinder of three senior Eskom employees. Section 24N(8) of NEMA, which provides for directors’ liability for environmental harm, did not extend to ordinary employees merely because of their seniority. No relief was granted against the cited Ministers or the National Energy Regulator of South Africa, as the necessary basis for relief against them had not been established. Topigs was ordered to pay the employees’ costs. Eskom was ordered to pay Topigs’ High Court costs and the reasonable costs of investigating and preparing the matter, including its expert’s qualifying fees. The successful portion of the appeal also carried a costs award in Topigs’ favour.

Concluding Remarks

The judgment highlights that environmental compliance requires effective implementation and enforcement of the conditions under which a project is authorised. Operators should ensure that remedial plans are implemented and that monitoring complies with the applicable standards. Furthermore, it is apparent that a court may decline continuing judicial supervision while still granting mandatory orders capable of securing compliance. The appropriateness of a structural interdict must therefore be considered separately from the entitlement to other effective relief.

The post Kusile Pollution: SCA Orders Eskom and Regulators to Act appeared first on Werksmans Attorneys.

Artificial intelligence (AI) is ubiquitous, and the gambling and betting sector is no exception. AI is increasingly embedded in gambling and betting products and platforms. The question is not whether to engage with AI, but how to develop, deploy, use and regulate it effectively and responsibly. The policy and regulatory choices made now will influence whether AI is ultimately used to protect consumers or to exploit them.

For gambling operators, AI presents significant opportunities, from processing live information and developing advertising campaigns to creating new features, applications and games. But it also introduces regulatory challenges and risks. AI could support player safety, but it also has the potential to intensify harmful gambling behaviours.

Algorithms allow operators to understand player behaviour and preferences with growing precision. This enables tailored content, services and advertising, while potentially improving the player experience, reducing choice overload, improving stickiness and reducing churn.

AI can also be a powerful responsible-gaming tool, particularly online. Several studies have used player-tracking data and machine-learning models to predict problem gambling using behavioural and monetary indicators. The results have been promising and can inform targeted interventions, including personalised approaches to problem gambling.

Research in France illustrates how this can work. In 2015 and 2016, the French National Gambling Authority, then known as ARJEL, now ANJ, partnered with researchers to develop an online gambling harm-detection system using demographic and behavioural data. The resulting machine-learning models proved highly effective at classifying people by risk level, demonstrating how such research can inform harm-reduction strategies for online gambling.

AI remains an imperfect and developing technology. A key challenge is whether AI systems treat people fairly. AI can perpetuate and exaggerate human biases and inequities, including those based on socio-demographic characteristics. It can hallucinate. Its performance can change over time because of “drift”. It can also simply get things wrong.

AI raises distinct concerns in the gambling context. For example, research has found that large language models repeatedly made irrational, high-risk betting decisions in simulated gambling environments. Because AI systems learn from human-generated data, they can inherit cognitive distortions such as the gambler’s fallacy.

All of this creates a challenge for regulators. South African gambling legislation was drafted for a bricks-and-mortar world. South Africa does have laws protecting consumers in areas including data protection, consumer protection, competition, copyright and cyber law. But it does not yet have AI-specific laws of general application.

Internationally, different approaches are beginning to emerge. The EU AI Act establishes a risk-based classification framework for AI systems. It does not single out gambling, but AI systems used in gambling that fall within high- or limited-risk categories will trigger obligations.

The UK has adopted a cross-sectoral, principles-based approach rather than a single horizontal AI Act. Individual sector regulators, such as the UK Gambling Commission, must apply AI governance principles within their respective domains.

Malta is particularly proactive on this issue. Following consultations earlier this year, on 18 September 2026 the Malta Gaming Authority launched its AI Gaming Charter. The Charter is a sector-specific framework promoting the ethical and responsible use of AI across gaming platforms. It provides a voluntary, principles-based framework to support the responsible, transparent and proportionate use of AI across gaming operations, and practical guidance tailored to the realities of the gaming sector.

In line with the risk-based approach adopted in the EU, the Charter applies a risk-based approach, in which the level of documentation, testing, governance, transparency, human oversight and monitoring applied to an AI system should reflect the nature of the use case and its potential impact on players, player rights, gaming integrity, or regulated outcomes.

The Charter sets out the following principles, which are mapped to the EU AI Act and GDPR, and which are intended to respond directly to the practices, gaps and emerging risks observed across the gaming sector:

• Transparency, accountability and explainability
• Fairness and non-discrimination
• Environmental sustainability
• Data protection, security and governance
• Human oversight and responsibility
• Reliability, safety and robustness
• Compliance, monitoring and AI governance
• Reporting and metrics
• Notification requirements
• Involvement of the Malta Gaming Authority and the Malta Digital Innovation Authority

Existing gambling laws in South Africa and across Africa were not designed for AI. The regulatory gap is real and growing, and the faster AI develops, the harder that gap will be to close.

The post AI and gambling, the regulatory gap is growing appeared first on Werksmans Attorneys.