The short version

AI hallucinations could lead South African lawyers into misconduct claims. Former Mpumalanga premier Dr Matthews Phosa warned that fake case law generated by AI must not be trusted in court.

What happened

At the LexisNexis Future of Law Conference in Johannesburg, Dr Matthews Phosa, a former Mpumalanga premier and director at Frans Schutte & Mathews Phosa Inc, warned that AI hallucinations, fabricated legal information produced by artificial intelligence, pose a serious risk to the legal profession.

He said presenting invented case law to a court could expose lawyers to accusations of dishonesty and professional misconduct.

Presenting fabricated legal information to a court could expose practitioners to accusations of dishonesty and professional misconduct.

Dr Matthews Phosa

The background

South Africa has seen two AI‑hallucination cases: in April, the Communications and Digital Technologies Department withdrew a draft National AI Policy after finding up to six AI‑generated false references, and the Home Affairs Department suspended two officials when fabricated references appeared in a white paper on citizenship and immigration.

Phosa highlighted that AI can generate convincing but false legal precedents, academic journals and other citations, and urged lawyers to verify any AI output before using it in legal memoranda or court submissions.

What it means for owners and operators

Legal firms and compliance teams must treat AI as a research aid, not a source of truth, and put processes in place to double‑check AI‑generated case law, citations and arguments before filing documents.

Failing to verify AI output could lead to disciplinary action, reputational damage and costly lawsuits, directly affecting profit margins and client trust.

Firms that embed verification steps, training and oversight will gain a competitive edge by delivering faster research while avoiding the risk of hallucinations.

Our take

Analysis by Infercline

AI hallucinations can expose lawyers to misconduct claims, verify every AI‑generated citation.

What to do in your business tomorrow

  1. Schedule a short training session on AI hallucinations for all legal staff this week.
  2. Audit recent court submissions for any AI‑derived content and flag any inaccuracies.
  3. Assign a senior partner to periodically review AI‑assisted research outputs for accuracy and compliance.

Related coverage

Sources

  • itweb.co.za: AI must not replace human judgement in SA courts
  • The Citizen: Phosa warns AI must never replace human judgment in pursuit of justice
  • MSN: AI can aid justice, not replace human judgement: CM
  • India Today: CJI Surya Kant urges tech-driven courts without replacing human judgement
  • Newsweek: AI Must Strengthen, Not Replace, Human Judgment | Opinion