NOTE

Focus on responsibility rather than disclosure: AI ethics in South African legal practice

Author: Donrich Thaldar

ISSN: 1996-2177
Affiliations: Distinguished Professor, University of South Africa;
Honorary Professor of Law, University of KwaZulu-Natal
Source: South African Law Journal, Volume 143 Issue 4, p. 658-675
https://doi.org/10.47348/SALJ/v143/i4a2

Share

Cite this article

Donrich Thaldar
Focus on responsibility rather than disclosure: AI ethics in South African legal practice
South African Law Journal, Volume 143 Issue 4, p. 658-675
https://doi.org/10.47348/SALJ/v143/i4a2

Abstract

A recent proposal by Van Eck (2025 TSAR 454) would amend the Code of Conduct for Legal Practitioners to require prior disclosure to, and prior consent of, the client before any use of generative artificial intelligence (‘AI’) in legal practice. The case against that proposal is set out in this note: it mischaracterises the international landscape, has no support in South African jurisprudence on the duty of honesty and integrity, cannot be enforced as a policy, and would chill precisely the use of generative AI that s 34 of the Constitution most plainly favours. An affirmative responsibility-centred alternative is developed, drawing on the recently revised ethics guidelines for legal practitioners on the use of generative AI; three reactive carve-outs are identified to address the cases that the contrary view is right to worry about.