From left: Antony Makins, chairman of the Special Interest Group for Artificial Intelligence and Robotics, IITPSA; Dr Machiniba Sylvia Sathekge, CIO, National Health Laboratory Service; Dr Xolile Sibande, information and cyber security expert; and Nkosikhona Nkosi, CIO, CCMA. (Image: Strikapose Studios)
Building trust in AI within SA’s public sector will depend on stronger governance, clearer accountability and improved skills development, particularly as adoption remains limited.
This emerged during a panel discussion at the ITWeb AI Summit 2026 in Bryanston, moderated by Antony Makins, chairman of the Special Interest Group for Artificial Intelligence and Robotics at the IITPSA. Panellists were Nkosikhona Nkosi, CIO at the CCMA; Dr Machiniba Sylvia Sathekge, CIO at the NHLS; and Dr Xolile Sibande, information and cyber security expert.
Makins noted the discussion came shortly after the introduction of SA’s AI Policy Framework. “We were only introduced to the policy framework a couple of weeks ago. Some of the questions are not easy,” he said.
Nkosi said AI adoption in the public sector remains limited due to gaps in governance. Trust in AI systems is tied to transparency and accountability, particularly where automated decisions affect citizens.
“AI in the public sector is still scarce in South Africa,” he said, adding that many users cannot challenge or understand automated outcomes. “There must be explainability and a human in the loop.”
He also noted fears around job security. “There is a distinct fear that people have about AI taking jobs.”
Sathekge said the new AI Policy Framework creates an opportunity for broader participation. “We are asked to comment on our own AI voice and make the AI policy that we want.” She emphasised that governance should enable innovation, citing the policy’s flexible, risk-based approach.
She advocated for testing AI systems before deployment. “Let’s do it, let’s fail fast and have a sandboxing approach.”
Drawing on her work in healthcare, she stressed the consequences of AI errors. “Behind that data point are people. We conduct 450 000 tests a day; yes, AI could do it faster, but if it gets it wrong, that could affect so many lives.”
On accountability, Sibande said responsibility must rest with deploying organisations, not vendors. “Legal accountability lies with the institution that is deploying the solution and not the service provider.” She added that citizens need to know which decisions are taken or informed by AI systems.
Skills development was identified as a priority. Sibande cautioning against focusing only on advanced expertise without building foundational knowledge. “There’s no point in focusing on expertise when the basics aren’t there.” She suggested government should consider centres of excellence and partnerships with academia and the private sector to address skills shortages.
In closing, Nkosi warned of risks including bias and foreign-developed systems that do not reflect local realities. “Taking externally or foreign-built solutions without local context is one of the challenges,” he said, noting that even basic tools struggle with local language and speech patterns. He also questioned how accountability should be enforced. “How do we ensure there’s accountability right up to the minister?”
The panel concluded that governance, ethics and skills development will be critical to responsible AI adoption in SA’s public sector.
