The department of communications & digital technologies in April published an AI planning discussion document aimed at initiating discussion between the public and private sectors to facilitate innovation and development in artificial intelligence.

The document drew mixed reviews, with some saying it was riddled with errors and unfinished thoughts, while others felt it to be a good first step in the policy process. A key criticism raised was its failure to mention or link to government’s draft national data and cloud policy.

This month, the department released an AI policy framework document, and is seeking stakeholder input as the basis for the subsequent establishment of a full-fledged national AI policy. It is anticipated that such a policy will inform the eventual promulgation of AI-related legislation.

The framework recognises the need to balance the effective and beneficial exploitation of the benefits of AI against ethical, social and economic imperatives. It emphasises the importance of “human-centred AI”, which augments and incorporates human decision-making, as opposed to replacing it. Further, the need for capacity building and economic development is tempered by an appreciation that AI must be harnessed ethically and responsibly to promote inclusive growth across South Africa.

The framework aims to address three key elements:

The quantitative “push of the present” factors such as rapid technological advancement, economic necessities and social demands, and a need to harmonise South African policy with international standards;
The “pull of the future” as set out in the National Development Plan, including a vision of sustainable economic transformation and development, social equity, and the aim of establishing South Africa as a global and regional leader in AI; and
The obstacles and entrenched structures referred to as the “weight of the past”, including the persistent digital divide, historical socioeconomic inequities, institutional inertia and resistance to change, and the need to update and reform existing regulatory frameworks, which are ill-equipped to keep up with the rapid pace of technological change.

The framework sets out various fundamental strategic pillars upon which an AI policy may be based:

Talent development and capacity development: To ensure that South Africa has a robust AI talent pool, the framework proposes incorporating AI into educational curricula, developing specialised training programmes and fostering partnerships between academia and industry.
Digital infrastructure: To create an environment that is conducive to AI innovation, the framework proposes the development of supercomputing infrastructure and digital connectivity infrastructure and technologies.
Research, development and innovation: To advance technological capabilities, the framework proposes establishing dedicated research centres, promoting public-private partnerships and providing funding and incentives for AI research and start-ups.
Public sector implementation: Aiming to enhance government efficiency, the framework proposes implementing AI to optimise state management and service delivery, and to develop complementary guidelines for ethical and effective deployment of AI in government operations.
Ethical AI guidelines development: The framework proposes the development of wider guidelines to ensure responsible and ethical AI use, addressing issues such as bias, fairness, transparency and accountability. These objectives would be achieved not only through guidelines, but also through regulatory compliance and governance policies.
Privacy and data protection: To ensure the safeguarding of personal information, the framework sets out three key objectives: to establish standardised data generation and utilisation practices across the public and private sectors, to strengthen data protection regulations, and to ensure transparency in AI data usage and storage practices.
Safety and security: To protect citizens and infrastructure, the framework proposes the implementation of robust cybersecurity protocols and the development of risk management measures.
Transparency and explainability: Recognising the need to build public trust in AI, the framework proposes the implementation of public awareness campaigns to educate the public on AI technologies and their implications. To achieve transparency, the framework proposes ensuring clear and open operation of AI systems, and making AI input and decision-making processes and outcomes understandable and accessible to users and stakeholders. Further, the framework aims to develop trust, accountability, and bias detection and mitigation by promoting AI systems that provide clear and easily understandable outputs.
Fairness and mitigating bias: By ensuring that AI systems are trained on inclusive and diverse data sets and developing methods to identify and mitigate biases in AI systems, the framework proposes the promotion of equitable AI deployment.
Human-centred approach: To maintain human oversight over AI, the framework proposes ensuring that critical AI decisions, particularly in generative AI, involve human oversight.
Professional responsibility: To foster the responsible use of AI, the framework proposes the creation of a code of conduct for AI professionals and the integration of ethical training into AI education and professional development.
Promotion of cultural and human values: To align the development of AI with key societal values, the framework proposes that AI systems be developed to promote human well-being, equality and environmental sustainability. This would require the involvement of diverse stakeholders in the policy-making process.

The framework is intended to serve as a strategic blueprint for the harnessing of AI technologies. With a focus on ethical guidelines, it aims to promote economic growth, technological advancement and societal well-being in line with South Africa’s values and priorities.

While AI has not hitherto appeared to be a priority feature for the South African government, the publication of the planning discussion document in April and the framework this month suggests a shift in this approach.

In addition, the department of communications’ newly appointed minister, Solly Malatsi, may seek to drive broader AI implementation and regulation (based on the Democratic Alliance’s 2024 election manifesto, which included a commitment to employing AI technologies to fight cybercrime and embrace modern technological solutions to service delivery challenges, and promote science, technology, engineering and mathematics education, for example).

While it has been reported that the communications department is seeking stakeholder input on the framework, the manner and timeline of this process are not yet clear.

The authors are Daniel Pretorius, partner, John Paul Ongeso, senior associate, Claire Franklyn, consultant, and Aneesa Valodia, candidate legal practitioner, Bowmans South Africa