Kieswetter Showcases AI-Powered Efficiencies in SARS at G20 Summit

The South African Revenue Service (Sars) has achieved remarkable success in its tax collection initiatives, a feat largely credited to the implementation of artificial intelligence (AI) and the lack of excessively stringent regulations surrounding new technologies.

“We are charting new territory, which requires a careful balance between policy development and practical application,” stated Sars Commissioner Edward Kieswetter during a roundtable discussion on Thursday. The event included representatives from central banks, the Financial Sector Conduct Authority (FSCA), and the Organisation for Economic Co-operation and Development (OECD) regarding the influence of AI in financial systems.

Read:
Sars recovers nearly R8 billion in tax debt in May
Sars monitors bank account activities
G20 finance leaders face uncertainty during Trump’s trade conflict
G20 works to finalize a communiqué, according to South Africa’s finance chief

This dialogue was part of a side event at the G20 Finance Track taking place in Zimbali, KwaZulu-Natal.

“If policy advances faster than practical implementation, we risk inhibiting experimentation and the exploration of emerging technologies,” Kieswetter cautioned.

He explained that Sars has already utilized machine learning and AI to assess 65% of taxpayers—just nine days into the tax filing season—without these individuals needing to submit a tax return.

“We relied entirely on third-party data, employing machine learning, algorithms, and AI for assessments and to evaluate fraud risk.”

Two decades ago, Sars needed six months to assess six million taxpayers.

“This year, two million taxpayers opting to file will receive assessments in five seconds.”

In the first nine days of the 2025 tax season, Sars issued R12 billion in refunds while collecting R3.5 billion through assessments.

Kieswetter emphasized that AI also helped Sars avert R5.5 billion in improper refunds.

‘Emerging and complex risks’

Opening the roundtable discussion, South African Reserve Bank Governor Lesetja Kganyago warned that while AI offers significant potential for innovation and financial inclusion, it also introduces new and complex risks that challenge supervision.

“Our mandate is to safeguard financial stability, ensure efficient market operations, and provide consumer protection. These core commitments are tangible promises to our constituents.”

Kganyago highlighted that AI presents “a notable duality – a landscape rich with opportunity intertwined with substantial risk.”

On one hand, it can improve risk detection at both institutional and systemic levels, enhance consumer credit assessments, and expand inclusion.

On the other hand, these same tools can amplify market shocks and perpetuate biases.

Kganyago stressed that regulators must stay true to their core mandates while adapting to technological changes.

“Technology will evolve. The tools we use will change significantly. However, our mission will remain the same – to ensure stability in the value of the money held by those engaging with the financial system.”

AI integration in South Africa’s financial sector

Katherine Gibson, divisional head at the FSCA, presented new findings from an upcoming joint survey conducted with the Prudential Authority.

The survey covered nearly 2,000 institutions across banking, payments, insurance, fintech, and other sectors, evaluating AI utilization in South Africa’s financial industry.

“The survey suggests a sector that is cautiously optimistic about AI’s capabilities, with varying levels of adoption,” remarked Gibson.

In 2024, AI investments in the global financial sector surpassed $50 billion (R892.7 billion), with projections to exceed $150 billion (R2.7 trillion) annually by 2028.

Locally, fintech firms are leading AI adoption, with two-thirds already utilizing it. Over 50% of banks have joined in, while the insurance and lending sectors are comparatively conservative, with adoption rates below 10%.

“Banks are investing nearly 20 times more than other sectors.”

However, she emphasized that innovation carries responsibilities.

“We must ensure that hyper-personalization doesn’t lead to hyper-exploitation and that generative AI chatbots do not mis-sell complex products to uninformed consumers.”

She cautioned that some institutions lack transparency measures.

“About one-fifth of South African institutions reported not using any explainability models or methods for their AI systems, raising concerns regarding transparency, accountability, and consumer protection.”

For many institutions, South Africa’s data protection law (Popia) is seen as the main barrier to AI implementation. However, Gibson challenges the notion that regulation hinders progress.

“These are not impediments to innovation; they are crucial guardrails ensuring that technology serves the public good.”

Read: FSCA progresses in enhancing the financial services sector
Listen: FSCA plans to invest R200 million to bolster its regulatory capacity

OECD highlights risks of market instability

OECD Secretary-General Mathias Cormann cautioned that generative AI poses increasing threats of financial fraud, scams, and cyber attacks, particularly as global financial markets become more interconnected through similar AI technologies.

“The application of AI in financial markets could escalate market volatility for many participants relying on similar AI models, leading to a convergence of trading and feedback loops,” Cormann warned.

The OECD’s latest risk monitor indicates that AI used in credit assessments and insurance underwriting could compromise privacy and lead to discrimination. Cormann also mentioned that competition, rather than concentration, is vital for leveraging the advantages of AI in finance.

Follow Moneyweb’s comprehensive finance and business updates on WhatsApp here.