The recent withdrawal of South Africa’s draft National Artificial Intelligence Policy presents an important opportunity to strengthen the country’s approach to AI governance through greater transparency, scientific rigour, expert oversight and public trust.

This is the view of the Academy of Science of South Africa (ASSAf), which said on Tuesday that while the withdrawal has raised concerns regarding the integrity, transparency and reliability of policy development processes, it also creates an opportunity to build a more robust, consultative and evidence-informed framework for AI governance.

The Department of Communications and Digital Technologies withdrew its draft National Artificial Intelligence Policy in April after it emerged that it contained various fictitious sources in its reference list.

Communications and Digital Technologies Minister Solly Malatsi, at the time, said that the failure compromised the integrity and credibility of the draft policy.

The draft AI policy was approved by Cabinet on March 25, and published in the government gazette on April 10 for public comment.

ASSAf said that this highlighted the growing importance of transparency in the use of AI-generated content and the need for appropriate human oversight in verifying and validating such outputs.

While AI can support decision-making and can be powerful tool in research and policymaking, it cannot replace scientific judgement or expertise or substitute for institutional integrity or human accountability.

“AI is increasingly capable of synthesising information, identifying patterns and extending analytical capacity. However, AI systems do not participate in the scientific and social processes through which knowledge is tested, debated and validated.

“Where evidence informs public policy, rigorous review and validation remain essential,” said ASSAf.

ASSAf emphasised that the ethical responsibility for ensuring the quality, reliability and integrity of evidence is shared by researchers, policymakers and institutions, and, as policy decisions affect public resources and access to essential services, weaknesses in evidence or oversight can have significant societal consequences.

Following the release of the Draft National AI Policy, ASSAf’s AI Steering Committee highlighted the need for stronger conceptual clarity, analytical coherence and evidence-informed policymaking.

The committee stressed that AI governance frameworks should balance innovation and implementation with safeguards related to human rights, equity, societal wellbeing, data governance and digital sovereignty.

The committee also identified the need to address fragmentation across institutions and sectors while strengthening AI literacy, critical engagement and public understanding across schools, universities, government and broader society.

“ASSAf intends to continue contributing scholarly and evidence-based perspectives to national discussions on AI through expert engagement, evidence synthesis and independent review processes.”

ASSAf’s AI Steering Committee has identified several priorities that should inform future AI policy development, including transparency and human oversight; integrity and reliability of evidence; responsible use of AI in policymaking; expert review and accountability; AI literacy and capacity development; coordination and coherence; communication and public trust; and epistemic justice and African knowledge systems.

The transparency and human oversight priority points to the appropriate disclosure of AI use in policy development processes and continued reliance on human expertise and judgement to validate AI-generated outputs.

The integrity and reliability of evidence priority establishes strong standards for evidence quality, traceability and verification and the use of reliable and fit-for-purpose AI models and datasets.

Meanwhile, the responsible use of AI in policymaking includes the verification and clear identification of AI-generated content and the documentation of methodologies and information sources used in policy development, while expert review and accountability delivers independent expert scrutiny of AI-informed policy outputs and clear accountability mechanisms for decisions informed by AI.

The priorities of AI literacy and capacity development and coordination and coherence would ensure, respectively, expanded AI literacy and critical engagement capabilities across government, research institutions, higher education, schools and society, and improved alignment across institutions, sectors and governance structures while recognising differing capacities and contexts.

Communication and public trust will ensure transparent communication regarding uncertainties, limitations and risks associated with AI systems, while epistemic justice and African knowledge systems recognise the risk that AI systems may reproduce existing inequalities in knowledge production and provide support for African languages, African knowledge systems, local datasets and scholarly contributions from the Global South to ensure inclusive participation in AI-driven innovation and policymaking.

Overall, it is an opportunity to build a stronger national AI framework, ASSAf said.

“Engagement with AI is rapidly becoming an essential life skill and a critical component of digital literacy. A successful national AI policy should clearly define the responsibilities of government, academia, industry and civil society in ensuring that AI serves the public good.

“The withdrawal of the Draft National AI Policy should not be viewed as a setback, but as an opportunity to develop a stronger, more credible and more inclusive policy framework,” ASSAf said.

The academy cautioned that without a rigorous and evidence-informed process, South Africa risks undermining its ability to achieve national development priorities, including those articulated in the National Development Plan and the Science, Technology and Innovation Decadal Plan.

“ASSAf therefore calls for a strengthened policy process that prioritises transparency, expert oversight, inclusive consultation and evidence-based decision-making to ensure that artificial intelligence contributes positively to South Africa’s social and economic development.”