The Chief Justice has issued the Judiciary (Artificial Intelligence) Practice Directions, Kenya’s first binding court rules governing the use of AI tools in judicial proceedings. They apply to every advocate, litigant in person, judge, and judicial officer across all courts and tribunals under the Judiciary’s superintendence, from the Supreme Court to Kadhis Courts, and they are in force now. Unlike the Draft AI Policy, which sets out institutional governance principles, the Practice Directions impose enforceable obligations with real consequences for non-compliance: struck-out documents, personal costs orders against advocates, referral to professional disciplinary bodies, and potentially contempt of court. At their core, the Practice Directions rest on three obligations that every user of AI in a court context must understand: disclose that AI was used, verify every citation and legal proposition independently, and take full professional responsibility for everything filed.
In practice, the disclosure obligation requires advocates to attach a prescribed AI Disclosure Statement (Annex A) to every court document in whose preparation any AI tool was used, identifying the tool, describing the nature of the assistance, and confirming independent review. On citations, the Practice Directions establish a strict liability standard: reliance on an unverified AI-generated citation constitutes a breach of the duty of accuracy regardless of whether the advocate was aware of the error, given that AI tools are expressly noted to generate fictitious, inaccurate, or misattributed references. Courts may enquire at any hearing whether authorities have been independently verified, and may decline to consider any that cannot be confirmed. On AI-generated evidence, parties wishing to tender AI-generated content must file a formal notice (Annex C) disclosing the tool used, the prompt given, and any modifications made, and bear the burden of proving authenticity and relevance. Judges and judicial officers face their own set of rules, including an express prohibition on using AI-generated summaries as a substitute for reading the actual record.
Two tensions in the Practice Directions deserve attention before they are finalised. The first is a conflict with the Draft AI Policy, which prohibits the use of third-party AI tools to predict outcomes in ongoing proceedings, a prohibition that sits uneasily with the Practice Directions’ permission for AI-assisted preparation of submissions, and that is broader than its stated purpose requires, given that outcome analysis over public judgments is a foundational function of legal practice. The second is a gap in scope: the Practice Directions apply only to courts and tribunals under the Judiciary’s superintendence and say nothing about AI use in alternative dispute resolution, mediation, or arbitration, creating an unregulated parallel channel in the justice system. These are issues we raised in our submissions to the Law Society of Kenya, and we encourage the legal profession to engage with the review process, which is set to take place within twelve months of the Practice Directions’ issue. In the meantime, every advocate practising in Kenya should treat AI disclosure as a standard element of every court filing and ensure that no citation goes to court unverified.
Read our full analysis of the Judiciary (Artificial Intelligence) Practice Directions, including section-by-section commentary on disclosure obligations, the strict liability standard for citations, the evidentiary framework for AI-generated content, and our observations on gaps requiring harmonisation. Find the detailed blog below.