When the Drafter is Algorithmic: Rethinking Pleading Rules in the Age of Artificial Intelligence

Introduction

The integration of Artificial Intelligence (AI) into contemporary legal practice has revolutionised how legal professionals approach their work.¹ AI offers opportunities for efficiency and access to justice through its key applications such as legal research and document review.² On the downside, the rapid adoption of AI raises significant concerns about ethical implications, legal accountability, and the potential for systemic bias.³ These challenges necessitate the crafting of a balanced legal framework that fosters innovation while safeguarding fundamental legal principles.⁴ Within this broader debate, one particularly pressing issue is the use of AI tools in drafting pleadings as evident from the case underpinning this commentary.⁵ This emerging practice presents distinct legal and ethical concerns, yet the Kenyan legal system has yet to craft a regulatory framework to govern the use of AI in legal drafting.⁶ This regulatory gap may lead to judicial decisions that resolve the immediate dispute but fall short of providing a forward-looking approach to the responsible integration of AI in Kenya’s legal system.

The 2025 case of Savla v Commission on Administrative Justice ended in favour of the applicant.⁷ Justice J. Chigiti declared that the Commission on Administrative Justice had breached the applicant’s rights to equality, dignity, information, and fair administrative action.⁸ The court also ordered the Commission to provide a comprehensive schedule detailing the status of all the applicants’ complaints within two months and awarded him the costs of the suit.⁹ The respondents appealed the decision on the grounds that their failure to file a response arose due to confusion occasioned by multiple applications filed by the respondent within a very short period of time.¹⁰ In addition, the appellants alleged that the respondents’ pleadings were generated using AI as a ground for challenging the suit.¹¹ In April 2026, Justice J. Chigiti reversed his earlier judgment after it emerged that the respondent, Nayan Savla, had illegally relied on machine-generated pleadings.¹²

Paragraph eighty-one of the judgment is particularly significant, as the judge stated that using artificial intelligence to generate pleadings may give one party an unfair advantage, thereby limiting the opposing party’s access to justice. The judge further argued that personalised AI drafting tools, being outside the existing rules of drafting, depart from established practice and risk creating a litigation disaster through the lack of uniformity.

The court’s objection to AI-assisted pleadings

The drafting of Pleadings in Kenya is regulated by Order 2 of the Civil Procedure Rules. The essence of order 2 is that it regulates the substance and format of pleadings. Crucially, rule 15 allows for the striking out of pleadings with a safeguard provided in rule 14, prohibiting technical objections based on mere want of form.¹³ As expressly mentioned in the judgment, the rationale for having these rules of drafting is to create a universal standard that all litigants and lawyers adopt when engaging with the courts in Kenya.¹⁴ The uniformity in drafting is argued to ensure that all litigants have fair access to justice . A departure from these rules by the use of unknown tools or AI is argued as creating a litigation disaster, leaving the judges with no guidance during proceedings.¹⁵ Another argument against the use of AI in drafting pleadings is that it may confer an unfair advantage on the user over other litigants. Its use has been likened to bypassing the established rules governing the drafting of pleadings, thereby undermining fairness in the litigation process. The merits of the arguments against AI in drafting are discussed in the following section.

Critiquing the decision

Artificial intelligence and the silence of procedural rules

The need for uniformity in drafting cannot be emphasised enough; however, this should not be taken to mean that artificial intelligence has no place in legal practice. Instead, the judgment highlights the need to ensure that the adoption of AI does not alter established procedural safeguards or undermine the fairness and equal access to justice that those safeguards are intended to protect. This concern must be understood within the broader context of the legal profession’s long-standing technological evolution. Historically, the legal profession has evolved from the analogue stage to the computational stage characterized by the emergence of AI and its application to law.¹⁶ AI-powered tools are helping lawyers to save valuable time by automating tedious, time-consuming tasks such as document review, legal research, and contract drafting.¹⁷ In addition, lawyers rely on AI technology to draft documents,¹⁸ the systems backed by AI are capable of quickly producing accurate and high-quality work.¹⁹

The present decision appears to reflect a degree of reluctance towards the incorporation of artificial intelligence in legal drafting, particularly where its use may disrupt established procedural norms. This reluctance is notable given that the AI-generated pleadings themselves did not appear to contain any hallucinations or substantive inaccuracies; in that regard, a stronger basis for criticism may have been the respondent’s failure to disclose the use of AI in their preparation. However, this should not be mistaken for institutional resistance. The draft Judiciary AI Practice Directions shows a move towards disclosure requirements as opposed to outright rejection of AI in legal practice.²⁰ The prevailing judicial approach seems to be one of regulated integration rather than outright rejection. The question that remains is how to achieve the effective integration of AI technology while maintaining trust and confidence, minimising risk of error and bias, and ensuring sufficient training across the legal profession.²¹

The concerns raised about the use of AI in the legal profession are valid and need to be addressed. The absence of comprehensive regulations governing the use of AI technology in the legal profession raises concerns about the ethical use of AI.²² Judge Chigiti rightly noted that drafting mechanisms assisted by AI are not expressly contemplated under the Civil Procedure Rules.²³ However, the absence of a regulatory provision presents an invitation to consider the development of  guidelines and practice directions on the ethical use of AI in legal practice and a further considerations of implications of its use where it creates discrepancies. Responding effectively to these legal and ethical challenges will require a collaborative approach involving legal professionals, technologists, policymakers, and ethicists.²⁴ Crucially, legal practitioners are expected to continually build AI literacy through ongoing training, and they must adhere to ethical and professional standards when using AI tools.²⁵ A starting point for the legal field to consider, as it develops regulations that protect the ethics of the practice in light of AI use, are the pillars of privacy, confidentiality, accountability, and supervision.²⁶ Privacy and confidentiality require informed consent, strong data protection measures, and systems to detect and address security risks.²⁷ Accountability and supervision ensure that lawyers are trained to oversee AI, that responsibility for its use is clearly assigned, and that outputs are transparent and traceable.²⁸

The Kenyan court system has, in recent years, demonstrated a clear capacity to adopt technological innovations that enhance the efficiency of the litigation process. This has been demonstrated through the integration of virtual courts, the electronic filing system, and the case tracking system.²⁹ Building on this trajectory, the regulated integration of AI technologies represents a logical and progressive next step, with the potential to further improve access to justice and strengthen overall judicial efficiency.

Common concerns regarding AI use include the potential for the spread of inaccurate information through AI hallucinations.³⁰  Hallucinations refer to a phenomenon whereby AI systems generate plausible but inaccurate or fabricated responses.³¹ In legal matters where precision is paramount,  inconsistency with legal doctrines and judicial precedents as a result of AI hallucinations may lead to wrong legal advice or decisions.³² Courts have, in some instances, struck out or questioned decisions due to the inclusion of AI-generated hallucinations. In Mata v. Avianca, Inc., the plaintiff’s counsel cited six fictitious judicial opinions that had been produced by ChatGPT.³³ When the court challenged these citations, the submitting attorneys attempted to conceal their methodology, which prompted the judge to impose sanctions for conduct undertaken in bad faith.³⁴ It is important to note that the present case did not identify any instance of hallucinations. The commission could not point to any fabricated citations or false cases.³⁵

Drafting of pleadings using AI gives the user an unfair advantage over the other litigants

The most controversial part of the judgment was the argument that the use of AI to draft pleadings gave the applicant an unfair advantage over the Commission.³⁶ The judges’ concern was that the use of AI represented a deviation from the uniform standards required of all litigants. The judge explicitly described the applicant’s personalised, machine-driven drafting methods as deplorable because they allowed the applicant to bypass the nature of legal documentation required by the rules, while his opponent had to adhere to the stated rules.³⁷ The notion of unfair advantage in this judgment is better understood as one litigant being permitted to deviate from established rules of practice while the other adheres strictly to them. This has been misconstrued in public discourse as the use of AI tools inherently making one party more capable than the other.

The merits of this argument have already been discussed above and include the need for uniformity and the assurance that all litigants enjoy the equal benefit of the law as espoused under Article 27 of the Constitution. This argument can however be challenged to the extent that it ignores the fact that the use of AI for a self-represented litigant would serve as a levelling tool that facilitates the right to access justice under Article 48 of the Constitution. The high cost of lawyers and the complexity of the law has historically locked people out of justice, consequently, this judicial deficit has perpetuated poverty and compromised fundamental rights.³⁸ AI systems can help bridge this gap by simplifying complex legal doctrines, thereby reducing structural barriers to understanding the law without a lawyer and enabling more accessible, lower-cost legal assistance.³⁹ Proponents of AI often frame it as a powerful equaliser in a society marked by deep inequalities,⁴⁰ particularly within the legal profession, where legal fees and the complexity of the law create significant barriers to access. Properly deployed, AI is not intended to replace lawyers, but to expand access to justice by supporting self-represented litigants and helping laypersons navigate complex legal issues.⁴¹ However, its use must be approached with caution, especially given the risks posed by AI-generated inaccuracies or hallucinations.

The question of strict compliance with set rules of procedure is of central importance to the argument. The judge argued that the court could neither tell the technology used to draft the pleadings nor the commands that the applicant gave to the computer to generate the results that he generated through AI.⁴² This section of the judgment significantly contrasts with what Order 2 of the Civil Procedure Rules envisioned. Order 2 rule 3(1) focuses on the final output and does not mandate the litigant to explain their drafting process, research methodology, or the specific software used.⁴³ Rule 4 provides an explicit list of matters that a party is legally required to plead specifically, yet none of the judge’s requirements fall within these rules.⁴⁴ Rule 16 provides that a pleading is perfected by the signature of an advocate or the party suing.⁴⁵ The law only regulates the final commitment to the facts, not the process. The human drafter remains solely responsible for the sworn facts. In the judgment, the applicant had verified the pleadings and argued personal responsibility for all factual statements on oath.⁴⁶ On this basis, the judges’ reasoning appears to conflate the uniformity of an applicable standard and that of the method used to meet the standard. The decision raises questions as to the presumed additional obligations to litigants which are not prescribed under the rules.

David Engstrom posits that rules in the context of legal proceedings should be treated as protective guardrails rather than barriers.⁴⁷ Arguably, the strike-out order can be interpreted as more of a barrier as it does not mandate correcting the AI-assisted pleading or require verification of its contents; it simply eliminates the litigant’s access to the court on the basis of the tool used, not the quality of the output.

As courts continue to evaluate the appropriate role of AI in judicial processes, they can start by implementing pilot projects such as encouraging controlled experiments in low-stakes cases, to compare AI-assisted motions with traditional drafting.⁴⁸ Alongside these practical trials, regulatory efforts should focus on reinforcing practitioners’ professional duty to verify, edit, and ultimately take responsibility for AI-generated outputs. A lawyer is mandated to always review the work product of a generative AI just as the lawyer must do so for the work of non-lawyer assistants, such as paralegals.⁴⁹

Beyond professional legal practice, AI also presents an opportunity to reduce legal costs and improve access to justice. To tackle the issue of legal fees as a barrier to court access, self-represented litigants should be encouraged to leverage and guide computer-driven legal assistance.⁵⁰ With appropriate regulation and clear institutional guidance, AI therefore has significant potential to enhance both the accessibility and effectiveness of Kenyan courts.

Conclusion

The decision in Savla v Commission on Administrative Justice represents one of the earliest judicial confrontations in Kenya with the use of artificial intelligence in litigation. While the court’s concerns regarding fairness, procedural uniformity, and the risk of misuse are legitimate, its approach ultimately conflates the method of drafting with the legal sufficiency of the final pleading. The Civil Procedure Rules regulate what pleadings must contain and how parties assume responsibility for them; they do not prescribe the technological tools through which those pleadings are prepared. This raises an important policy question for lawmakers as to whether the rules should be reviewed in light of the growing use of artificial intelligence in litigation. In this regard, a key research question emerges: whether the current procedural framework adequately accommodates the use of AI in legal drafting, or whether it reveals gaps that require reform. Against this background, this commentary has argued that the absence of explicit procedural authorisation for AI should not be construed as a prohibition. Rather, it ought to be understood as a regulatory gap that calls for deliberate reform.

Closely supervised AI has the potential to reduce litigation costs, improve efficiency, and expand access to justice. The appropriate judicial response, therefore, is not exclusion, but regulation: clear professional guidelines, mandatory human oversight, and procedural safeguards capable of detecting inaccuracies and preventing abuse. In this way, procedural rules can serve as guardrails that preserve the integrity of the legal process while allowing Kenya’s legal system to evolve with technological change.

Image was generated using Adobe Firefly.