DAR ES SALAAM: ARTIFICIAL intelligence has become an ind ispenable tool in dispute resolution. It can assist with legal research, document review, translation, transcription, case management, and the preparation of draft submissions. These technologies improve efficiency, reduce costs, and make arbitration more accessible. However, when AI begins to influence substantive aspects of an arbitral process particularly the evaluation of evidence, legal reasoning, or the drafting of awards significant due-process and enforceability concerns arise.
rom the perspective of judicial practice, the greatest concern is that AI must never replace independent human judgment. The legitimacy of arbitration rests on the parties’ confidence that their dispute has been determined by impartial arbitrators exercising their own reasoning. If an arbitrator delegates critical functions to AI, parties may legitimately question whether they have received a fair hearing or whether the award genuinely reflects the tribunal’s own assessment of the evidence and applicable law.
Such concerns directly implicate fundamental procedural guarantees, including the right to be heard, transparency in decisionmaking, and the tribunal’s duty to independently evaluate the issues before it. Where AI influences substantive reasoning without appropriate human oversight, an arbitral award may be vulnerable to challenges on the basis that the tribunal failed to exercise its own judgment or that the parties were denied a fair opportunity to address matters that influenced the outcome.
We have already seen examples demonstrating the risks associated with uncritical reliance on generative AI. In FAM v ZAM, High Court of Tanzania at Mbeya, in Matrimonial Appeal No. 212 of 2025, counsel cited six non-existent authorities generated by AI tools. The High Court emphasised that advocates owe duties to the court, their clients and the administration of justice to accurately represent the law, verify authorities, and use technology competently. The Court further observed that misuse of AI may attract sanctions, including adverse costs orders, striking out proceedings, referral to professional disciplinary bodies, contempt proceedings, or even criminal liability.
Similar concerns have emerged internationally. Courts in Kenya, South Africa and Zimbabwe have all encountered fabricated AI-generated citations, reinforcing that technological innovation cannot substitute for professional competence or ethical responsibility. Although these cases arose in litigation, the lessons are equally applicable to arbitration. An arbitral award may
face challenges to recognition or enforcement if parties can demonstrate that AI compromised fundamental procedural guarantees, including the right to be heard or the tribunal’s independent exercise of judgment. AI therefore remains a valuable assistant, but it cannot become the decision-maker.
While AI can greatly enhance the efficiency of arbitration, certain functions must remain exclusively within the authority of arbitrators because they require legal reasoning, judgment and accountability that technology cannot replicate.
These core adjudicative functions include:
• assessing the credibility of witnesses and weighing evidence;
• determining disputed facts;
interpreting and applying the law;
• deciding procedural fairness and the admissibility of evidence;
• exercising discretion throughout the proceedings; and
• making and signing the final award with reasons that genuinely reflect the tribunal’s own independent analysis.
AI may appropriately assist with administrative and preparatory tasks such as organising documents, translating materials, preparing chronologies, summarising evidence, conducting legal research, or generating preliminary drafts but it should never make decisions that require judicial discretion or determine the outcome of the dispute.
To safeguard due process and preserve the enforceability of arbitral awards, courts, arbitral institutions and tribunals should adopt clear safeguards.
• First, there should be institutional guidelines governing the responsible use of AI in arbitration, clearly identifying both permissible and prohibited uses.
Secondly, arbitrators and counsel must independently verify the accuracy of all AI-generated legal authorities, factual summaries and translations before relying on them.
• Thirdly, where AI has played a material role in preparing submissions or tribunal documents, appropriate disclosure should be considered where necessary to preserve transparency and maintain the parties’ confidence in the process. •
Fourthly, all confidential information entered into AI systems must be adequately protected to safeguard privacy, confidentiality and legal privilege. •
Finally, continuous training should be provided to arbitrators, counsel and judges so that they understand both the capabilities and the limitations of AI and can exercise effective human oversight.
ALSO READ: AG stresses negotiations skills
In Tanzania, the Judiciary has already taken important steps towards the responsible adoption of technology through e-filing, virtual hearings, digitisation of court records, AIassisted transcription and translation systems, and legal research tools such as Tanzlii. It has also developed training materials on artificial intelligence for judges and magistrates, including in partnership with UNESCO, recognising that technological innovation must always be accompanied by ethical safeguards and judicial oversight.
Ultimately, in conclusion AI should remain a tool that supports, rather than replaces, human decision-making. Arbitration derives its legitimacy not from technology, but from the confidence that disputes are resolved by independent, impartial and accountable arbitrators. Efficiency should never come at the expense of fairness, and the continued evolution of AI must be matched by safeguards that preserve due process, equality of the parties, and the integrity of the arbitral process.