Standard Bank And Stellenbosch University Build New Responsible AI Framework As South African Businesses Race Into Artificial Intelligence

JD GLOBAL MEDIA | SOUTH AFRICA

21 SEPTEMBER 2026

JOHANNESBURG — A new partnership between Standard Bank and Stellenbosch University is putting responsible artificial intelligence at the centre of South Africa’s rapidly expanding AI economy, with the two institutions developing a framework designed to guide how businesses develop, deploy and use increasingly powerful AI systems.

The collaboration brings together financial-sector experience, data science, computational thinking and ethics to address one of the most important questions emerging as artificial intelligence moves from experimentation into everyday business operations: how can companies use AI at scale while maintaining security, accountability, privacy and meaningful human oversight?

The partnership has produced the SBSA Responsible AI Approach, a framework intended to establish practical parameters for the responsible development and use of artificial intelligence.

The initiative arrives at a time when South African businesses are moving beyond simply testing generative AI tools and beginning to integrate AI into customer service, financial analysis, fraud detection, software development, decision-making and internal operations.

For companies in sectors such as banking, insurance and financial services, the stakes are particularly high because AI-generated decisions can potentially affect people's finances, access to services and business opportunities.

The new framework therefore focuses not only on what AI can do, but also on how organisations should manage the risks associated with deploying it.

AI IS MOVING FROM EXPERIMENT TO BUSINESS INFRASTRUCTURE

Artificial intelligence has rapidly moved beyond being a technology demonstrated primarily through chatbots and experimental applications.

Businesses are increasingly using AI to analyse information, automate repetitive tasks, identify patterns and support employees.

Generative AI has also made sophisticated technology available to employees who may not have specialist programming knowledge.

A worker can ask an AI system to summarise a document.

A developer can use AI to assist with software development.

A customer-service operation can use AI to analyse conversations.

A financial institution can use machine-learning systems to identify unusual transaction patterns.

A company can use AI to analyse large amounts of information that would take human teams considerably longer to process.

These capabilities create opportunities for productivity and innovation.

They also introduce new risks.

AI systems can produce inaccurate information.

They can reproduce biases contained in their training data.

They can expose confidential information if poorly managed.

Automated decisions can be difficult to explain.

Employees may become overly dependent on machine-generated recommendations.

And organisations may struggle to determine who is responsible when an AI-supported decision produces an unexpected outcome.

These concerns form part of the reason Standard Bank and Stellenbosch University have focused their collaboration on responsible AI.

A FRAMEWORK FOR RESPONSIBLE AI

The SBSA Responsible AI Approach is designed to provide a structure for developing, deploying and using AI systems responsibly.

The framework focuses on principles including safety, robustness, security, human oversight and accountability.

The intention is to create a foundation that businesses can use when introducing AI into their operations.

Rather than treating ethics as an issue to consider after an AI system has already been deployed, the framework places responsible practices throughout the AI lifecycle.

That means organisations need to think about responsibility when an AI system is being designed.

They need to consider it when the system is developed.

They need controls when it is deployed.

And they need continuing oversight once the technology is operating.

This approach recognises that AI systems can change over time.

Models can be updated.

Data can change.

Business processes can change.

New risks can emerge.

A system that was considered appropriate when introduced may require additional controls later.

WHY THE BANKING SECTOR NEEDS STRONG AI CONTROLS

Financial services is one of the sectors where artificial intelligence can have particularly significant consequences.

Banks handle sensitive financial information.

They make decisions that affect customers and businesses.

They operate within extensive regulatory requirements.

They also have to maintain trust.

AI can potentially improve financial services in several ways.

It can assist with fraud detection.

It can help identify unusual transactions.

It can support customer-service employees.

It can improve the speed at which information is analysed.

It can assist with financial education.

It can help businesses understand customer behaviour.

It can support onboarding and other administrative processes.

But these same applications create questions about how automated systems reach conclusions.

If an AI system contributes to a financial decision, organisations need to understand what information influenced that decision.

If a customer is affected by an automated process, there needs to be appropriate accountability.

If sensitive information is processed by an AI system, the organisation must ensure that the information is protected.

This is why responsible AI has become a major issue for financial institutions.

HUMANS REMAIN PART OF THE PROCESS

A major principle behind the new framework is the importance of human oversight.

AI can process information at enormous speed.

It can detect patterns across datasets.

It can automate repetitive tasks.

But it does not eliminate the need for human judgement.

People can consider context.

They can understand relationships.

They can recognise when an unusual situation requires additional investigation.

They can make ethical judgements.

They can communicate with customers.

They can take responsibility for decisions.

The framework therefore does not treat AI as a complete replacement for human decision-making.

Instead, the objective is to create systems where technology can increase capability while people remain responsible for appropriate oversight.

This approach is particularly relevant when AI is used in high-impact areas.

AI AND THE QUESTION OF TRUST

Trust has become one of the central issues surrounding artificial intelligence.

Consumers may use AI without knowing exactly how an answer was generated.

Employees may rely on AI-generated recommendations without checking every detail.

Businesses may introduce AI tools supplied by third parties without fully understanding the underlying models.

This can create a gap between the apparent confidence of an AI system and the reliability of its output.

A system can provide an answer that sounds convincing while still being incorrect.

In business environments, such errors can have consequences.

An incorrect summary could result in a misunderstanding.

An inaccurate analysis could influence a business decision.

A flawed recommendation could affect a customer.

A security failure could expose sensitive information.

Responsible AI therefore requires organisations to understand the limitations of the technology rather than simply focusing on its capabilities.

BIAS AND DISCRIMINATION

Another major concern is the potential for bias and discrimination.

AI systems learn patterns from data.

If the information used to develop or operate a system contains historical biases, an AI system can potentially reproduce or amplify them.

This is particularly important when AI is involved in decisions affecting people.

A business needs to understand whether its systems produce consistently different outcomes for different groups.

It also needs processes for identifying and addressing unintended bias.

The problem is not always obvious.

An AI model can produce statistically accurate results in some circumstances while still producing unfair outcomes in particular cases.

This makes ongoing monitoring important.

Responsible AI is therefore not simply about checking an algorithm once.

It involves continuously evaluating how the system behaves in the real world.

TRANSPARENCY BECOMES MORE IMPORTANT

AI systems can also create transparency challenges.

Traditional software usually follows rules that programmers have explicitly written.

Modern AI systems can operate differently.

Some models contain enormous numbers of parameters and generate outputs based on complex patterns.

This can make it difficult for ordinary users to understand exactly why a particular answer or recommendation was produced.

Businesses therefore need to determine what level of explanation is appropriate for different applications.

A low-risk internal productivity tool may require a different level of transparency from an AI system contributing to a decision affecting a customer's finances.

The higher the potential impact, the more important appropriate explanation and oversight become.

PRIVACY AND AI

Privacy is another major issue addressed by the responsible AI conversation.

AI systems can process enormous quantities of information.

In a banking environment, this can include highly sensitive financial and customer information.

The ability to process information does not automatically mean that every piece of information should be used.

Businesses need to determine what information is necessary for a particular AI application.

They also need to ensure that information is handled securely.

Privacy protections become especially important when AI systems are connected to large organisational databases.

An employee using an AI tool to summarise information, for example, needs to understand what happens to the information entered into the system.

If sensitive information is transferred to an external platform without appropriate controls, the organisation could face significant security and compliance risks.

SECURITY MUST DEVELOP ALONGSIDE AI

Artificial intelligence can help companies strengthen cybersecurity, but AI systems themselves also need to be protected.

Businesses are increasingly using AI to detect unusual behaviour and identify potential threats.

At the same time, criminals can use AI to create more convincing phishing messages, automate attacks and develop new methods of social engineering.

This creates a continuing technological race.

Companies need to protect their AI systems, their data and the infrastructure supporting them.

They also need to consider what could happen if an AI system is manipulated.

Security therefore forms part of responsible AI rather than being a separate issue.

AI GOVERNANCE CANNOT BE STATIC

One of the important points emerging from the partnership is that responsible AI frameworks cannot simply be written once and left unchanged.

Artificial intelligence continues to evolve.

New models are released.

New capabilities appear.

New applications are developed.

New vulnerabilities are discovered.

Regulations can change.

Public expectations can change.

The Stellenbosch University School for Data Science and Computational Thinking has therefore emphasised that the framework should remain evolving and adaptable.

This is significant because an AI governance framework that works for today's technology may not necessarily be sufficient for future systems.

Organisations need processes for reviewing and updating their controls.

THE ROLE OF UNIVERSITIES IN AI GOVERNANCE

The partnership also illustrates the growing role universities can play in shaping the development of artificial intelligence.

Universities bring expertise in areas such as data science, mathematics, computer science, ethics and research methodology.

Businesses bring practical experience from deploying technology in real-world environments.

Combining the two can allow researchers and industry specialists to examine problems from different perspectives.

In this case, the collaboration brings together Standard Bank's experience in financial services and business technology with Stellenbosch University's expertise in data science, computational thinking and ethics.

The result is intended to be more practical than a purely theoretical discussion of AI ethics.

FROM PRINCIPLES TO PRACTICAL APPLICATION

One challenge facing responsible AI initiatives is translating broad principles into everyday business decisions.

It is relatively easy to say that an AI system should be fair, secure and accountable.

The difficult part is determining what those principles mean when an organisation is building an actual product.

For example, a development team may need to decide what data can be used to train a model.

A risk team may need to determine what level of human review is required.

A security team may need to establish access controls.

A legal team may need to examine regulatory requirements.

Business managers may need to decide whether the benefits of automation justify the risks.

Employees may need training so they understand how to use AI responsibly.

A useful framework must therefore connect high-level principles with practical processes.

STANDARD BANK IS ALREADY EXPANDING ITS AI OPERATIONS

The responsible AI framework comes as Standard Bank continues to increase its use of artificial intelligence.

The bank has been developing internal AI capabilities and deploying tools designed to help employees work with generative AI.

One example is FRAME, an internal customer-product development platform designed to accelerate how teams create and test customer solutions.

The bank has said that some early use cases reduced the time between an idea and a live solution by as much as 70%.

The organisation has also been developing AI-agent capabilities.

One example involves a voice-recognition system designed to connect customers with a banker who has the relevant skills for their specific issue.

The bank has said that this system is expected to reduce transferred calls and customer frustration.

These applications illustrate why governance becomes important as AI moves from experiments into operational systems.

When AI starts affecting real customer interactions, the organisation needs clear controls.

AI AGENTS ARE CHANGING THE CONVERSATION

The emergence of AI agents adds another dimension to responsible AI.

Traditional AI tools often wait for a person to ask a question.

Agentic systems can potentially perform a sequence of actions to accomplish a task.

This can make AI more useful.

It can also increase risk.

An AI system that only generates text may produce an incorrect answer.

An AI system capable of taking actions can potentially make an incorrect decision or trigger an unintended process.

As organisations deploy more agentic technology, governance frameworks will need to address not only what an AI system says but also what it is allowed to do.

Permissions, monitoring, escalation mechanisms and human intervention become increasingly important.

DATA QUALITY WILL SHAPE AI RESULTS

Another major issue is the quality of the data used by AI systems.

Artificial intelligence can be extremely sophisticated, but poor data can undermine its usefulness.

Incomplete information can produce incomplete results.

Outdated information can produce outdated recommendations.

Inconsistent information can create unreliable patterns.

Sensitive information can create privacy risks.

For financial institutions, strong data governance is therefore essential.

Standard Bank has been investing in its data infrastructure and moving critical data into modern platforms to support its AI ambitions.

The responsible AI framework complements that technological investment by addressing how the resulting AI systems should be used.

THE ECONOMIC IMPACT OF RESPONSIBLE AI

AI has the potential to affect South Africa's economy well beyond individual companies.

Businesses that use AI effectively may be able to automate processes, improve productivity and develop new services.

Employees may use AI tools to handle repetitive work and focus more time on tasks requiring judgement and creativity.

New companies can build products around artificial intelligence.

Universities can train specialists.

The technology sector can develop new infrastructure.

But the economic benefits will depend on trust.

Customers are unlikely to embrace AI applications if they believe their information is unsafe or decisions are unfair.

Businesses may hesitate to invest if regulatory uncertainty becomes too high.

Responsible AI can therefore become an economic issue as well as a technology issue.

SOUTH AFRICA'S DEVELOPING AI LANDSCAPE

The partnership comes as South Africa continues to develop its broader artificial intelligence capabilities.

Businesses, universities and government institutions are increasingly examining how AI can be used to improve services and productivity.

At the same time, policymakers are working on questions surrounding AI governance, digital sovereignty, data and emerging technologies.

The growth of AI creates opportunities for South Africa to develop local expertise rather than simply importing technology developed elsewhere.

Responsible AI frameworks can form part of that ecosystem.

They can help businesses develop confidence around the use of AI while providing structures for managing risks.

THE IMPORTANCE OF LOCAL CONTEXT

AI systems are often developed for global markets.

But the way technology affects people can vary between countries.

South Africa has its own legal environment, languages, economic conditions and social context.

A system that works well in another country may require adaptation before it can be deployed locally.

This is particularly relevant in financial services.

South African customers may have different financial behaviours from customers elsewhere.

Local businesses may have different needs.

Regulatory requirements can differ.

Language and cultural context can influence customer interactions.

Developing responsible AI practices locally can therefore help ensure that technology is evaluated within the environment where it will actually be used.

TRAINING PEOPLE FOR THE AI ECONOMY

Responsible AI also requires skilled people.

Employees need to understand how AI works at a basic level.

They need to know what the technology can and cannot do.

They need to understand when human verification is necessary.

They need to recognise potentially inaccurate or biased outputs.

They need to understand how confidential information should be handled.

This means AI adoption is not simply an IT project.

It can require organisational change.

Managers, legal teams, compliance officers, data specialists, developers and ordinary employees can all have roles to play.

A FRAMEWORK THAT CAN EVOLVE

The emphasis on an adaptable framework is particularly important because artificial intelligence is changing faster than many traditional technology governance systems.

A company could develop a policy for one generation of AI and discover that a newer generation introduces capabilities that the policy never anticipated.

For this reason, responsible AI governance needs regular review.

Organisations may need to update policies when new models become available.

They may need to reassess risks when AI systems gain new capabilities.

They may need to change controls when new regulations are introduced.

They may also need to learn from real-world incidents.

A framework that can evolve is therefore more likely to remain useful as technology changes.

WHAT THIS COULD MEAN FOR SOUTH AFRICAN BUSINESSES

The Standard Bank and Stellenbosch University initiative could provide lessons for other South African organisations considering AI adoption.

Businesses do not necessarily need to wait until something goes wrong before creating governance structures.

They can establish principles before deploying high-impact systems.

They can define responsibilities.

They can determine which decisions require human involvement.

They can identify sensitive data.

They can establish security controls.

They can create processes for investigating AI failures.

They can train employees.

And they can regularly review whether the controls remain effective.

This approach allows businesses to pursue AI innovation while recognising that technological capability and organisational responsibility need to develop together.

THE NEXT PHASE OF AI IN SOUTH AFRICA

The next phase of artificial intelligence in South Africa is likely to involve more than chatbots and productivity tools.

AI is increasingly moving into customer service, financial services, healthcare, government, cybersecurity, manufacturing and other areas.

As that happens, the consequences of AI decisions will become more significant.

That makes responsible deployment increasingly important.

The partnership between Standard Bank and Stellenbosch University reflects this transition.

The focus is no longer simply on whether AI works.

The discussion is increasingly about how it should work, who should oversee it, how risks should be managed and how trust can be maintained.

A TECHNOLOGY CHALLENGE WITH A HUMAN DIMENSION

Artificial intelligence may be built from algorithms, data and computing infrastructure, but its consequences are ultimately human.

A customer affected by an automated decision is a person.

An employee working alongside an AI system is a person.

A business owner relying on an AI-generated analysis is making decisions that can affect other people.

A developer responsible for an AI application has obligations beyond simply making the system function.

This human dimension explains why responsible AI has become such an important part of the technology conversation.

SOUTH AFRICA ENTERS A MORE SERIOUS AI ERA

The new responsible AI framework developed by Standard Bank and Stellenbosch University arrives at a moment when artificial intelligence is becoming increasingly embedded in the South African economy.

The technology can create new opportunities for productivity, innovation and customer service.

But those opportunities come with responsibilities.

AI systems need to be secure.

They need appropriate human oversight.

Their use needs to be accountable.

Data needs to be protected.

Potential bias needs to be examined.

Customers and employees need appropriate safeguards.

And governance frameworks need to evolve alongside the technology.

The collaboration between a major financial institution and a leading academic data-science institution demonstrates how industry and academia can work together on these challenges.

The result is not a claim that AI can be made completely risk-free.

Instead, it establishes a structured approach for identifying and managing risks while allowing businesses to continue exploring what the technology can accomplish.

As South Africa's AI economy develops, the ability to combine innovation with responsible governance could become increasingly important.

The technology is moving quickly.

The frameworks governing it will have to keep moving too.

JD GLOBAL MEDIA will continue to follow South Africa’s artificial intelligence, digital transformation, cybersecurity and emerging technology developments.