Regulatory Updates on the use of AI in the legal profession

 source is Unsplash
 Image Source : Unsplash

Regulatory Updates on the Use of AI in the Legal Profession

Discussions on Artificial Intelligence (AI) in the legal profession in Africa have shifted to regulatory and governance action. There has been an increase in professional bodies and institutions moving to develop practice guidelines, compliance frameworks, or governance policies for legal practice, with jurisdictions such as Kenya, Nigeria and South Africa taking proactive steps. Aside from the developments in Africa, we look at Australia, where the courts and associations are making efforts to bridge regulatory gaps on the use of AI in legal practice.

Africa: Kenya, Nigeria and South Africa

Kenya

In Kenya, there is a draft AI policy for the Judiciary, which is yet to be finalised. The policy aims to modernise court administration while safeguarding fundamental rights. It addresses systemic operational challenges by leveraging AI to reduce case backlogs, automate court transcription, deploy virtual legal assistants for underserved communities and optimise case allocation. Under the draft policy, institutional oversight is vested in the AI Governance Committee (AIGC), which is constituted from the existing Integrated Case Management Systems (ICMS) Committee.

According to the policy, no AI system is to be deployed without AIGC authorisation, ensuring that third-party vendors and internal systems undergo rigorous technical, legal and ethical audits. There is a mandatory Judicial AI Impact Assessment (JAIIA) proposed to evaluate a tool’s potential impact on constitutional rights, data flows, privacy and human override mechanisms. The policy mandates prioritising culturally sensitive local-language interfaces like Kiswahili and requires judicial data to be stored and managed locally within Kenya.

The policy also establishes a risk classification matrix that categorises AI tools by severity, strictly prohibiting third-party systems from processing or editing confidential case files, bench notes, witness statements or undelivered judgements. It further institutionalises capacity building through the Kenya Judiciary Academy to integrate AI ethics and governance and ensure training is practical, role-specific and continuous to keep pace with technological advancements.

In In the Matter of the Estate of Jujhar Singh Didar Virdee & others (2026), a recent case in the High Court of Kisumu, the emerging use of AI in the legal profession was brought to the attention of the court. The Court used the case to illustrate that, while the use of AI in the legal profession is beneficial and efficient, there must be a duty of care and standard of professionalism maintained. The Court, on perusal of the Interested Party’s submissions, was unable to comprehend the arguments in the submissions. The Court stated that the submissions were disorganised and in an unusual format, prompting it to presume that the submissions might have been generated by an AI tool.

The court highlighted that while AI is useful in legal research and preparation of some court documents, it is necessary to verify the AI output to ensure accuracy. It further ruled that whenever AI is utilised it must be accompanied by human intervention, verification, editing and proper organisation. The ultimate responsibility for the accuracy, coherence and presentation of any document filed before a court rests squarely with the filing party or their legal representative, not the software or tool employed.

Nigeria

In Nigeria, the Bar Association, through its Section on Legal Practice, has published Guidelines for the Use of Artificial Intelligence in the Legal Profession in Nigeria. The Guidelines highlight the use of AI as a tool to enhance human legal practice. AI improves legal practice by automating repetitive work, supporting legal research, contract review, document management, predictive analysis and improving client service through document automation.

The application of AI must be within the scope of the Rules of Professional Conduct for Legal Practitioners 2023 (RPC). Legal practitioners must be aware of the limitations of AI tools. They must avoid feeding personal client information into the systems unless the AI tools have appropriate privacy mechanisms with appropriate settings in place. AI systems cannot be relied on to be accurate; thus, overreliance without proper understanding of the system’s shortcomings violates the duty of competence, caution and care to clients. Legal practitioners are required to exercise independent legal judgement and avoid presenting AI-generated content as their original work. They must protect evidence integrity and avoid conflicts of interest by checking AI tool ownership.

The Guidelines recommend mitigating algorithmic bias, requiring regular algorithmic fairness testing to prevent discriminatory legal outcomes. Lawyers are advised to carefully curate training data to ensure it is representative and unbiased when customising AI tools. The Guidelines also provide a formal AI Impact Assessment checklist that firms should execute prior to deployment. The checklist covers workflow impact, Nigeria Data Protection Act (NDPA) compliance, algorithmic bias, change management and shifts in lawyers’ roles.

The Guidelines also promote collaboration of practitioners with AI developers to ensure they cover their legal needs through technologies such as custom legal tools. This collaboration should include providing feedback to developers for system improvement and contractual clauses governing IP ownership. There must also be continuous AI literacy training to equip lawyers with skills for leveraging AI effectively in practice. The Guidelines further emphasise that the risks of AI systems must be managed and not just acknowledged.

South Africa

In South Africa, the regulatory stance governing the use of artificial intelligence in judicial proceedings is established through the interpretation of existing professional conduct rules and judicial precedent rather than a specific, standalone AI statute. Under the Code of Conduct for all Legal Practitioners, Candidate Legal Practitioners and Juristic Entities, practitioners have a mandatory obligation to take all reasonable steps to avoid misleading the court on matters of fact or law, which extends to ensuring that all written pleadings, affidavits, and cited authorities placed before a judge are genuine and accurate.

The judicial intolerance for fabricated AI research culminated in the landmark ruling of Philani Godfrey Mavundla v The MEC (2025). In that matter, the applicant’s legal representatives filed a supplementary notice of appeal relying on nine cited case authorities. Upon independent perusal by the court and its legal researchers, only two of the nine cases existed, and even those were cited inaccurately and concerned completely irrelevant legal issues, while the remaining authorities were non-existent, fabricated citations.

Demonstrating the inherent hazards of such technology, Bezuidenhout J tested ChatGPT with the cited case names and found that the system fabricated confirmations that the fictitious decisions were real. Declaring that relying on AI technologies for legal research is irresponsible and unprofessional, the Court ordered the law firm to pay the costs of the additional appearances and referred the judgment to the Legal Practice Council for disciplinary investigation.

Mavundla marks a shift from earlier judicial leniency seen in Parker v Forsyth N.O (2023), signalling that courts have lost patience with practitioners who fail to verify AI outputs. In Parker v Forsyth N.O (2023), the plaintiff’s counsel relied on unverified LLM results, therefore presenting fictitious case law authorities to the defendant’s counsel. The court merely struck out the AI-generated paragraphs and ordered the Plaintiff’s counsel to pay sixty percent of the Defendant’s counsel costs, but did not refer the judgement to the Legal Practice Council.

The case places South Africa alongside international jurisdictions that have sanctioned practitioners for hallucinated filings, such as Park v Kim (2024) in the United States and Zhang v Chen (2024) in Canada, where the courts reprimanded counsel for relying on AI without exercising their duty of care as officers of the court. In Park v Kim (2024), the attorney was referred to the Grievance Panel for investigation and ordered to disclose the misconduct to their client.

In Zhang v Chen (2024), the counsel was required to pay extra costs for expenses and wasted hearing days investigating the matter. The counsel was also directed to be up to date with Law Society Notices. Legal practitioners have a duty to their clients and the court to assert the accuracy of the authorities they present to the court, and workload pressure, fatigue, or ignorance of technology can never serve as a legal defence. This is a point of growing importance for anyone involved in litigation and dispute resolution.

Global Outlook

Federal Court of Australia

In Australia, courts have issued practice notes and rules governing AI use in the legal profession. For instance, the Federal Court of Australia issued the Use of Generative Artificial Intelligence Practice Note (GPN-AI). It provides that Generative AI (Gen AI) must be used appropriately with due care, and it must not adversely affect the administration of justice. Areas where caution is necessary are pleadings, written submissions, lists of documents and other documents or information lodged with or sent to the court.

There must be disclosure of Gen AI use where appropriate. This includes where Gen AI tools are used to summarise information which a witness relies on to make a statement, to create media to be presented to the court, and in any other manner that may affect admissibility of evidence. Disclosure should occur at the start of the document and say concisely where AI was used. Court users must be conscious of their legal and professional obligations when employing any form of Gen AI.

Federal Circuit and Family Court of Australia (FCFCOA)

The Federal Circuit and Family Court of Australia (FCFCOA) published a Practice Direction on the Use of Artificial Intelligence (PD-AI). The Practice Direction directs that court users must ensure that false or inaccurate information is not submitted to the court. It is the responsibility of AI users to ensure that any content generated is verified and accurate, and to adapt material to the Australian legal context. Court users must exercise extreme caution with evidence, ensuring affidavits reflect the witness’s words rather than AI drafting and that expert witnesses certify that they have read the Practice Direction.

AI must never be used to commit litigation abuse or perpetrate family violence. If a GenAI tool is capable of retaining information entered into it, court users must refrain from including confidential or sensitive information, especially if the tool is public. This is to avoid compromising personal safety, endangering asylum seekers in migration matters and the protection of children. It highlights that courts may respond to misuse of AI by disregarding material, making costs or other orders, or referring matters to a regulatory body.

Queensland Law Society

The Queensland Law Society also recently launched an AI selection and use checklist for legal practitioners. The checklist requires firms to evaluate whether AI vendor pricing is financially viable. Firms must also evaluate licensing models and opt for privately licensed models that are hosted in isolation on dedicated infrastructure to prevent data leakage. There must be a Privacy Impact Assessment performed to ensure that integrating AI does not affect the firm’s confidentiality perimeter, in line with the firm’s wider data protection and privacy obligations.

There must also be enforceable contractual terms that ensure complete data recovery and deletion upon contract termination with vendors. There must also be oversight to ensure AI tool usage does not expand beyond the necessary scope. Prompts, system queries and draft copies must be logged and preserved to establish document provenance in court. Internal policies and procedures must also be developed.

The checklist establishes that AI adoption cannot be treated as a mere technological procurement; rather, it demands an integrated risk management framework balancing stringent vendor contracts, structural oversight, thorough records management and active human supervision.

Conclusion

Across various jurisdictions, cases have illustrated that, in the absence of formal AI practice guidelines, courts will apply existing professional conduct rules in cases involving legal professionals’ use of AI. The effect is that professional uncertainty is starting to slowly narrow, thus highlighting to legal practitioners that AI governance, verification protocols and risk management cannot be put off until legislators catch up. Legal professional bodies and institutions have also tried to bridge the gap to ensure that the standard of professionalism is maintained in the legal profession in the wake of developments in artificial intelligence.

For guidance on AI governance and compliance in your practice or organisation, contact our team.

Sources

  1. Judiciary of Kenya, Artificial Intelligence (AI) Policy for the Judiciary of Kenya (draft).
  2. In the Matter of the Estate of Jujhar Singh Didar Virdee & others (2026), High Court of Kenya at Kisumu.
  3. Nigerian Bar Association Section on Legal Practice, Technology and Law Committee, Guidelines for the Use of Artificial Intelligence in the Legal Profession in Nigeria (2024).
  4. Legal Practice Council (South Africa), Code of Conduct for all Legal Practitioners, Candidate Legal Practitioners and Juristic Entities, Notice 168 of 2019, Government Gazette 42337.
  5. Mavundla v MEC: Department of Co-Operative Government and Traditional Affairs KwaZulu-Natal and Others (7940/2024P) [2025] ZAKZPHC 2 (8 January 2025).
  6. Parker v Forsyth NNO and Others (1585/20) [2023] ZAGPRD 1 (29 June 2023).
  7. Park v Kim, 91 F.4th 610 (2d Cir. 2024), No. 22-2057.
  8. Zhang v Chen, 2024 BCSC 285.
  9. Federal Court of Australia, Use of Generative Artificial Intelligence Practice Note (GPN-AI) (16 April 2026).
  10. Federal Circuit and Family Court of Australia, Practice Direction: Use of Artificial Intelligence (PD-AI) (29 May 2026).
  11. Queensland Law Society, AI Selection and Use Checklist for Legal Practitioners (February 2026).
Share to your feed on social media
en_USEnglish