\n<\/p>\n
South Africa\u2019s National Financial Ombuds Scheme says it is increasingly frustrated with complainants using artificial intelligence to generate lengthy submissions packed with incorrect legal references, adding significantly to the organization\u2019s workload without improving outcomes for complainants.<\/p>\n
CEO Reanna Steyn told radio station 702 that the organization is receiving complaints running to 150 or 200 pages that cite various acts, legislation and case law \u2014 much of it inaccurate. \u201cWe\u2019ve checked with a program. In one matter, it was 97% AI-generated. It\u2019s not written by a human, and I\u2019m not saying it\u2019s all wrong, but it\u2019s certainly not all right,\u201d she said.<\/p>\n
Instead of quickly identifying the core issue and requesting a response from the relevant financial institution, staff must now work through lengthy documents that cite nonexistent cases and incorrect legislation. \u201cSome of the references are wrong, the legislation that they quote is incorrect, and you have to now sift through that,\u201d Steyn said.<\/p>\n
She added that when complainants receive a response, they often reply with another lengthy AI-generated document, further prolonging the process. Asked how many cases in which complainants used AI had improved the prospects of a successful outcome, Steyn said \u201cnone, as far as I am aware.\u201d<\/p>\n
Her advice to complainants is straightforward. \u201cIf you do want to make use of AI, then please be so kind as to go and check the references and not just send us everything that\u2019s there because some of those things are just plain wrong,\u201d she said. \u201cAt least make sure it\u2019s not 150 pages, and double-check.\u201d<\/p>\n
The issue extends beyond consumer complaints. Two South African lawyers have faced consequences for submitting AI-generated legal citations that turned out to be fabricated. In September 2024, Judge Elsje-Marie Bezuidenhout found two falsified case citations in court papers and described the conduct as \u201cirresponsible and downright unprofessional,\u201d instructing the law firm to pay the costs of two associated court hearings.<\/p>\n
In July 2025, a junior advocate\u2019s use of the AI tool Legal Genius \u2014 which generated fictitious case law in an urgent application related to a disputed refining license \u2014 was referred to the Legal Practice Council, resulting in one of the first South African judgments to directly address the misuse of generative AI in legal documents.<\/p>\n
\u201cIt reflects a growing trend in local and international courts. Judges are warning legal professionals about the risks of relying on unverified AI-generated content,\u201d said Refilwe Motsoeneng, an associate at Michalsons Giles Inc. \u201cThis case reinforces the ethical duty to independently verify all legal sources before citing them. Even unintentional AI \u2018hallucinations\u2019 can cause reputational harm and lead to professional misconduct complaints.\u201d<\/p>\n<\/div>\n
\n