South African Appeal Tests Algorithmic Grant Decisions
South Africa's government appealed a January 23, 2025 High Court ruling that found regulations governing the Social Relief of Distress grant unconstitutional and invalid. The case concerns online-only applications and automated eligibility checks used by the South African Social Security Agency, according to the Centre for Applied Legal Studies, while the Supreme Court of Appeal has admitted the Global Center on AI Governance as a friend of the court.
South Africa's government has appealed a High Court ruling that found regulations governing the Social Relief of Distress (SRD) grant unconstitutional and invalid. The appeal brings the legality of digital-only applications and algorithmic eligibility decisions before the Supreme Court of Appeal, according to the Centre for Applied Legal Studies (CALS).
The Pretoria High Court issued its judgment on January 23, 2025, in litigation brought by the Institute for Economic Justice and #PayTheGrants against the Department of Social Development, the South African Social Security Agency (SASSA), and the Minister of Finance, CALS reports. The government appealed that decision.
Automated checks and access to benefits
According to CALS, SASSA sought to administer the SRD grant exclusively through digital applications and use an algorithmic verification system to assess eligibility. CALS reports that applicants could be excluded when automated decision-making identified income above R624, even where applicants were otherwise eligible for the benefit.
The Conversation reports that SASSA launched the SRD program in May 2020 to provide temporary assistance to people with no income or other grants. Its account describes a R370 monthly payment and states that regulations moving administration under the Social Assistance Act limited applications to online channels and, in many cases, treated automated decisions as definitive.
The case is not solely a question of model accuracy. The reported dispute concerns whether a public-benefits system can make consequential determinations through automated verification while limiting applicants' access to alternative application channels and effective recourse.
AI governance arguments enter the appeal
The Global Center on AI Governance, represented by CALS, has been admitted as amicus curiae, or friend of the court, in the appeal. CALS states that the organization will address the human-rights implications of algorithms, AI, and automated decision-making in social-grant administration, with emphasis on the right to social security.
The Conversation article was written by Mark Gaffley, who discloses that he is the Global Center on AI Governance's director of legal and operations. That disclosure is relevant because the organization is participating in the appeal.
For public-sector ML teams, comparable disputes underline that a technically automated eligibility workflow is also an administrative process. Industry experience with high-impact automated decisions repeatedly places importance on data-quality controls, clear decision explanations, accessible human review, appeal mechanisms, and monitoring for disparate outcomes. The legal outcome in this case could therefore be closely watched by organizations developing or procuring automated systems for welfare and other public services.
Key Points
- 1South Africa's grant appeal tests whether automated eligibility decisions and digital-only access satisfy constitutional requirements for social assistance.
- 2CALS reports that automated income verification could exclude otherwise eligible applicants, making data quality and contestability central operational concerns.
- 3Comparable public-sector automation programs commonly require human review, appeal paths, and disparate-impact monitoring alongside model performance measurement.
Scoring Rationale
The appeal concerns algorithmic decision-making in a high-impact public-benefits system, where errors can directly affect access to income support. It is particularly relevant to practitioners building or procuring automated eligibility, fraud-detection, and public-service workflows, although it remains a jurisdiction-specific court case rather than a broad technical release.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
