Authored By: Sthandiwe Bhengu
North-west University
INTRODUCTION
The legal profession is now on the front lines of regulatory surveillance rather than a bastion of client confidentiality due to the global duty to combat financial crime. Legal professionals are increasingly being labelled as “gatekeepers” under the Financial Action Task Force (FATF), whose main responsibility is to maintain the integrity of the financial system rather than to completely secure client secrets. This change is especially noticeable in South Africa, where trust accounts are being scrutinised more closely as possible means of money laundering due to recent institutional attempts to get off the FATF “grey list.” Although enforcing anti-money laundering (AML) laws is a legitimate governmental goal, the current legal framework presents an intolerable moral conundrum. This article contends that by forcing practitioners to act as de facto state informants through the “No Tipping Off” rule, the South African AML framework, mainly through the Financial Intelligence Centre Act (FICA), disproportionately jeopardises the constitutional sanctity of legal professional privilege (LPP). This article first looks at the statutory conflict between the common law and FICA; second, it explores the practicality of reporting obligations; finally, it assesses how new AI-driven legal analytics significantly change the definition of “knowing” facilitation of crime.
THE STATUTORY COLLISION: FICA, POCA, AND THE COMMON LAW
The Prevention of Organised Crime Act (POCA) and FICA have different mandates, which is where the issue starts. When someone “knows or ought reasonably to have known” that property is part of the proceeds of illegal activity, POCA makes it illegal to facilitate money laundering. This objective concept of “ought reasonably to have known” establishes a risky liability threshold for solicitors. FICA extends on this by categorising lawyers as “accountable institutions” with stringent reporting requirements under Section 29.
Section 37 of FICA contains the “analytical heart” of this dispute. While Section 37(2) aims to maintain LPP, Section 37(1) expressly supersedes any duty of secrecy or confidentiality. The South African courts have established that this protection is not unqualified, nevertheless. The Constitutional Court ruled in Thint (Pty) Ltd v. National Director of Public Prosecutions that LPP is a fundamental right required for the adversarial system, but it must be asserted by the right-holder (the client) or their agent. The problem stems from the fact that FICA’s Section 29(3) “No Tipping Off” (NTO) rule forbids the lawyer from revealing the existence of a report. As a result, a client might not know that their privilege is being violated, making it impossible for them to assert it.
In addition, the Western Cape High Court set a rigorous five-part test for privilege in A Company v. Commissioner for the South African Revenue Services, which includes the necessity that the communication not aid in the commission of a crime. FICA effectively narrows the scope of LPP to the point where the attorney-client relationship is no longer a “safe space” for legal assistance but rather a site of potential state-mandated betrayal by requiring reports on “suspicious” transactions, a criterion significantly lower than “proven” criminality.
THE OPERATIONAL DILEMMA
The NTO rule’s operational reality compels solicitors to play a dishonest dual role. In order to preserve the “illusion of regularity” in the client relationship, a practitioner who notices a questionable transaction must file a questionable Transaction Report (STR) with the Financial Intelligence Centre (FIC). This results in what has been called the “wolf in sheep’s clothing” situation, in which the lawyer continues to represent the client’s interests in public while secretly supporting the state’s investigation.
This is a clear transgression of the Legal Practice Act’s uberrimae fides (utmost good faith) ethical criterion. A lawyer is required by the Code of Conduct for Attorneys to prioritise the interests of their clients. However, the criminal consequences for “tipping off”—which include up to 15 years of imprisonment—effectively force practitioners to put state compliance ahead of client loyalty.
The Financial Intelligence Centre’s own assessment, which states that although the risk of money laundering in the legal sector is “high,” the number of STRs filed by attorneys remains disproportionately low, further exacerbates this operational gap. This suggests that legal practitioners are either actively choosing to protect their clients’ trust at the risk of regulatory non-compliance or are having difficulty identifying suspicious activities. One major shortcoming of the current AML system is the absence of a defined empirical foundation for reporting without undermining the “trust necessary for an effective legal defence.”
THE DIGITAL FRONTIER: AI AND THE REDEFINITION OF INTENT
An unresolved frontier in this argument is the rise of AI-driven legal analytics. The notion of “knowing or intentional” support of money laundering under Section 4 of POCA is changing fundamentally as law firms use automated AML screening techniques. Large databases can now be cross-referenced by automated algorithms to find “layering” or “integration” trends that a human practitioner might overlook.
It is nearly impossible to justify the “ought reasonably to have known” threshold under POCA if an AI technology identifies a transaction as high-risk and the practitioner fails to take action. However, the recent rulings in Northbound Processing (Pty) Ltd v. South African Diamond and Precious Metals Regulator and Mavundla v. MEC shows that practitioners are held firmly accountable for the results of their technologies. In Mavundla, the court established that technological reliance did not absolve professional negligence by referring practitioners to the Legal Practice Council for neglecting to verify AI-generated research.
When it comes to AML, this means that using “legal tech” raises the practitioner’s “burden of knowledge” rather than acting as a shield. The NTO rule is immediately activated if an AI tool used by a company detects a suspicion. This creates a “digital trap” in which the practitioner becomes an unwilling participant in a state probe at the touch of a button since the speed of AI detection surpasses the practitioner’s capacity to morally disengage from a situation without informing the client.
CONCLUSION AND RECOMMENDATIONS
In South Africa, the conflict between attorney-client privilege and AML enforcement is at a breaking point. Global financial standards are important, but the existing NTO rule and the lower threshold for “suspicious” reports violate Sections 14 (Privacy) and 35 (Fair Trial) of the Constitution by turning legal counsel into state whistleblowers.
In order to address this it is advised that the legislature implement a “Safe Harbour” policy for solicitors. With this innovation, practitioners should be able to leave a situation as soon as a suspicion arises without it being legally interpreted as “tipping off.” The Legal Practice Council must also provide explicit guidelines on AI governance, making it clear that although technology helps with detection, human practitioners are still the ultimate guardians of the “interests of justice” and must strike a balance between compliance and their non-delegable duty to the client. South Africa can only meet FATF requirements without compromising the moral core of the legal profession by setting precise, empirical reporting restrictions.
REFERENCE(S):
TABLE OF CASES
A Company and Others v Commissioner for the South African Revenue Services 2014 (4) SA 549 (WCC)
Bogoshi v Director, Office for Serious Economic Offences 1996 (1) SA 784 (A)
Mavundla v MEC: Department of Co-Operative Government and Traditional Affairs KwaZulu-Natal ZAKZPHC 2
Northbound Processing (Pty) Ltd v South African Diamond and Precious Metals Regulator (Johannesburg High Court, 2025)
Thint (Pty) Ltd v National Director of Public Prosecutions ZACC 13
TABLE OF LEGISLATION
Constitution of the Republic of South Africa, 1996
Financial Intelligence Centre Act 38 of 2001
Legal Practice Act 28 of 2014
Prevention of Organised Crime Act 121 of 1998
SECONDARY SOURCES
Financial Action Task Force, International Standards on Combating Money Laundering and the Financing of Terrorism & Proliferation (FATF, 2025)
Financial Intelligence Centre, Assessment of the Inherent Money Laundering and Terrorist Financing Risks: Legal Practitioners (FIC, 2024)
Financial Intelligence Centre, Guidance Note 4B on Reporting of Suspicious and Unusual Transactions (FIC, 2019)
Jana VLM, ‘Leveraging Artificial Intelligence to Combat Money Laundering and Related Crimes in the South African Banking Sector’ (2024) 27(1) Potchefstroom Electronic Law Journal 64
Law Society of South Africa, Guidance and Resource Document for a Risk Management and Compliance Programme (LSSA, 2025)
Layman K, ‘Confidentiality v justice: The effect of the No Tipping Off rule’ (2025) 1 African Student Law Journal 1
SA Legal Academy, ‘The Hallucinatory Reality: AI Governance & Liability for South African Law Firms’ (2025) https://legalacademy.co.za accessed 5 August 2026
