The Constitutional Court's judgment in Absa Bank Ltd and Another v Commissioner for the South African Revenue Service [2026] ZACC 15, handed down on 22 April 2026, is the first interpretation of the General Anti-Avoidance Rules (GAAR) in sections 80A to 80L of the Income Tax Act 58 of 1962 at this level since those provisions were introduced in 2006. The court dismissed Absa's appeal by 9 to 1, upholding additional assessments that re-characterised exempt preference dividends as taxable interest.
The transaction and the assessment
Between 2011 and 2015, Absa and its subsidiary United Towers subscribed for ZAR 1.9 billion in preference shares in PSIC Finance 3, a Macquarie Group vehicle. The funds flowed downstream into entities Absa said it knew nothing about. Those entities used the capital to acquire Brazilian government bonds whose interest yield was exempt under the South Africa-Brazil double tax agreement, and the exempt income was passed back up the chain to Absa as preference dividends. SARS re-characterised the dividends as taxable interest under the GAAR. Absa argued that it could not be a party to an arrangement of which it had no knowledge, and it had not personally obtained a tax benefit. The court rejected both arguments.
"Party" is now an objective test
Section 80L defines a party as one who "participates or takes part" in an arrangement. Absa argued that one cannot participate in something one does not know about. The majority disagreed. Majiedt J held that participation is an objective inquiry. The question is whether the taxpayer's conduct forms a constitutive link in the composite scheme, not whether the taxpayer knew how each downstream step operated. The legislature chose "participates or takes part" rather than the familiar "knows or ought reasonably to have known" found elsewhere in South African statute. The court relied on the breadth of "any" in "any party" and the inclusion of "indirectly" elsewhere in the GAAR. To require knowledge would allow upstream funders to immunise themselves by deliberately choosing not to know how their capital was deployed. Wilful blindness is precisely the mischief the 2006 amendments targeted.
SARS may tax a party who did not personally receive the tax benefit
Section 80B(1) empowers SARS to determine the tax consequences of an impermissible avoidance arrangement "for any party." Absa argued that section 80G(1), which requires "the party obtaining a tax benefit" to rebut the presumption of avoidance purpose, confines liability to that party. The court held that section 80G addresses the burden of proof, not the scope of liability. SARS may therefore assess any party. The court held that Absa had in any event obtained a tax benefit. The correct but-for test asks what the tax treatment would have been if the transaction were stripped of its avoidance features. Once the conduit chain and the Brazilian bond mechanism are removed, Absa's return was functionally indistinguishable from taxable interest on a loan.
Where the difficulty of assessing GAAR risk lies
The GAAR risk at both ends of the spectrum is clear. The middle is where the boundary of the analysis remains unsettled.
From the judgment, at the high end of the GAAR risk is where a funder was a single-investor SPV with no commercial substance and the funding passed through a conduit chain with yield-protection mechanisms that insulated the funder from equity risk. The funding was then used in cross-border transactions which turned interest deductible at the resident borrower's level into exempt dividend income at the funder's level through a Brazilian government bond swap designed to benefit from an exemption under the SA-Brazil DTA.
On the other end of the spectrum where GAAR risk is minimal is where preference shares (with gross-up clauses) are issued by an operating company or held by many investors in a diversified portfolio. The issuer has substance, the dividends are paid out of operating cash flow, and the funder's return reflects genuine equity risk (despite the gross-up).
Between the two ends sits a range of structures that are neither obvious Absa fact patterns nor clean operating-company funding. The defensibility of any specific structure turns on whether, on the correct but-for test, removing the avoidance features produces a different tax outcome. The greater the GAAR risk, the more critical it becomes that there is a demonstrable commercial reason for the specific step and the structure as a whole.
Practical consequences for funders
Funding documentation may need to address the allocation of GAAR risk contractually. Standard preference share subscription agreements do not usually contain freestanding GAAR indemnities. The contractual protection available to funders typically takes the form of tax adjustment mechanisms that are time-bound and subject to carve-outs which may limit its effectiveness in a GAAR context. The Absa litigation spanned more than a decade from the first transactions to the Constitutional Court's judgment. Funders, issuers and arrangers may need to consider the scope, duration and exclusions of tax protection clauses in new preference share agreements.
Three key insights from this case for future disputes
Three principles from the majority judgment will shape future GAAR disputes.
- The correct but-for test to determine tax benefit asks what the tax treatment would have been if the avoidance features were stripped away, not what would have happened had the arrangement not been entered into at all. In this case, the court accepted the re-characterisation where a return labelled as an exempt dividend was economically indistinguishable from interest on a loan once the conduit structure was disregarded.
- The presumption of purpose in section 80G(1) is now an objective test. A taxpayer must rebut it by reference to the relevant facts and circumstances reasonably considered, not by asserting a subjective commercial motivation. Internal memoranda recording a business rationale will not be decisive if the objective features of the arrangement point to avoidance.
- Section 80G(2) permits the GAAR to be applied to individual steps within a broader arrangement. The commercial purpose for the composite transaction cannot be used to disguise a step with a tax avoidance purpose.
The dissent of Judge Rogers remains available to be argued where the facts of a matter can be distinguished from the arrangement in the judgment. The practical difficulty for taxpayers is that distinguishing the facts may require the dispute to reach the SCA at minimum, and possibly the Constitutional Court, before clarity is obtained. For now, a GAAR analysis must be considered at the structuring stage rather than a risk to be managed only in the event of an assessment, and the cost of treating it otherwise has become considerably higher.
