The General Division of the High Court of Singapore has issued a landmark decision on the professional use of AI in litigation. In <span class="news-text_italic-underline">Tajudin bin Gulam Rasul and another v Suriaya bte Haja Mohideen [2025] SGHCR 33</span>, the court imposed a personal costs order on counsel who cited a non-existent case generated by a generative AI tool. The judgment underscores that lawyers bear a non-delegable duty to verify all authorities placed before the court and that improper reliance on AI will attract sanctions.
In September 2025, the General Division of the High Court of Singapore delivered a decision of significant importance for legal practice in the era of generative AI. The court addressed the professional responsibilities of advocates and solicitors when using AI-assisted tools for legal research and drafting and made clear that ultimate responsibility for accuracy rests squarely with counsel.
The proceedings arose out of Originating Claim 125 of 2025. After the defendant failed to file a notice of intention to contest, default judgment was entered against her. She subsequently applied in Summons 1240 to set aside that judgment.
Anticipating a potential defence of illegality under the <span class="news-text_italic-underline">Moneylenders Act 2008 (2020 Re Ed)</span>, the claimants’ counsel filed written submissions on 1 June 2025 citing an authority said to support the proposition that ad hoc loans between acquaintances do not amount to moneylending unless carried on as a business for profit. That authority was not included in the bundle of authorities.
When opposing counsel attempted to locate the case, it became apparent that it did not exist. The case name and citation had been fabricated by a generative AI tool used by a junior lawyer within the claimants’ firm.
Opposing counsel alerted the issue on 18 July 2025. On 21 July 2025, one day before the hearing, the claimants’ counsel filed amended submissions and a supplementary bundle with a replacement authority, without seeking leave of court. The matter was initially characterised as a clerical or typographical error. It was only during the hearing on 22 July 2025, when questioned directly by the court, that counsel acknowledged the authority was fictitious and that AI had been used in preparing the submissions.
Assistant Registrar Tan Yu Qing emphasised that advocates and solicitors are officers of the court and must uphold the highest standards of professional conduct. This includes the fundamental obligation to ensure that every authority cited is genuine, accurate and appropriate. The court held that the citation of a fictitious case is wholly impermissible regardless of whether it results from human error or reliance on AI.
The court observed that such conduct undermines confidence in the administration of justice and results in wasted judicial time and unnecessary costs for the opposing party. It also referred to existing judicial guidance on the use of generative AI by court users which makes clear that AI outputs must always be independently verified and treated only as an aid. Verification cannot be outsourced to another AI tool and responsibility remains personal and non-delegable.
Applying the three-stage test derived from <span class="news-text_italic-underline">Ridehalgh v Horsefield [1994] Ch 205</span> as adopted in Singapore jurisprudence, the court found:
The Court also expressed concern about the lack of candour shown, noting that full disclosure occurred only after sustained questioning.
The court ordered the claimants’ counsel to personally pay S$800 to the defendant representing the costs incurred as a result of the fictitious citation. Payment was required within 14 days. Both counsel were also directed to provide their clients with a copy of the decision, reflecting the court’s view of the seriousness of the misconduct.
This decision represents a clear inflection point in Singapore’s approach to AI and legal ethics. While the use of AI tools in legal practice is not prohibited, the judgment confirms that courts will not tolerate fabricated or unverified authorities and will impose personal sanctions where professional duties are breached.
The ruling has important practical implications. Lawyers must implement robust verification processes grounded in authoritative sources such as official law reports and statutory databases. Senior practitioners must actively supervise junior lawyers and support staff and cannot delegate responsibility for accuracy. Prompt and transparent disclosure of errors is critical and attempts to downplay or obscure mistakes will aggravate the court’s response.
<span class="news-text_italic-underline">Tajudin bin Gulam Rasul and another v Suriaya bte Haja Mohideen</span> stands as a cautionary precedent for Singapore legal practice. It confirms the High Court’s firm stance that reliance on AI does not dilute professional responsibility and that fictitious or AI-generated authorities will attract personal consequences. The judgment reinforces the foundational principle that truth and accuracy are indispensable to the administration of justice and to maintaining public confidence in the legal system.
