Benefits, Risks and Limits of Using Generative AI in Dispute Resolution: Takeaways From Recent Guidelines and Developments

Overview

In January 2025, the Hong Kong Government established the Consultation Group on LawTech Development ("Consultation Group") as a policy initiative in the Chief Executive’s 2024 Policy Address[1] , to research and promote the use of lawtech.[2] The formation of the Consultation Group shows the positive attitude of the Administration towards the use of generative artificial intelligence (AI) in dispute resolution, demonstrating the commitment of the Government to actively address the limitations of AI applicable in the legal industry. Stakeholders’ meetings held throughout the first half of 2025 further reflect the proactive stance of the legal sector in keeping pace with technological advancement.

On the other hand, recent Court decisions show that Courts have taken a relatively cautious approach noting the concerns of confidentiality, accuracy and reliability of the use of AI. This is also reflected in the first set of ‘Guidelines on the Use of Generative Artificial Intelligence for Judges and Judicial Officers and Support Staff of the Hong Kong Judiciary’ published in July 2024 ("2024 Judiciary Guidelines"). In particular, the Hong Kong Judiciary noted that "using generative AI for legal analysis is not recommended", unless there is a generative AI model with proven ability to protect confidential, restricted and private information, and adequate built-in checking and verification mechanism to ensure accuracy and reliability.[3]

This article explores the developments in the usage of generative AI in dispute resolution, its benefits and risks, as well as some recently published guidelines on the use of AI in law. It concludes with our take on the more effective ways to promote benevolent use of generative AI in dispute resolution.

Benefits and Developments

Generative AI has demonstrated its ability to streamline various aspects of dispute resolution: to conduct legal research and analyse data and case authorities efficiently, to summarise and interpret voluminous documents and information, to translate documents and transcribe hearings at a fraction of the cost of human stenographers.[4] Generative AI has been observed to have alleviated the burden of traditionally tedious tasks that are repetitive and resource-intensive, in aspects of time and cost, allowing legal professionals to focus on more substantial matters in a contentious case. In fact, the English Court has acknowledged AI to be a "powerful technology" and useful in litigation, for example, "to assist in the management of large disclosure exercises".[5]

In Hong Kong, the most notable example of the use of generative AI is the voice-to-text technology ("VTT") currently being piloted by the Judiciary. VTT has been under active development and use since December 2023 for real-time note-taking in hearings to facilitate judgment writing[6] and is expected to be fully set up in the Court buildings of the Court of Final Appeal and the High Court by the third quarter of 2025.[7]

It is also encouraging to see the introduction of more AI tools catered for use in the legal context. In February 2025, the Hong Kong Generative AI Research and Development Centre ("HKGAI"), which is a co-operative research centre led by the Hong Kong University of Science and Technology, developed the first local large language model based on DeepSeek technology.[8] Examples of HKGAI’s inventions include a generative AI application that supports legal drafting, bilingual translation, referencing and summarization, a tool which automatically generates legal analysis, and an application which quickly summarises discussion in meetings without the need of note-taking. The Department of Justice is participating in the pilot use one of the HKGAI AI applications in July 2025. Legal practitioners have also been actively exploring the development of AI-powered legal products for use in-house.

Similar tools are particularly beneficial to litigants in person, allowing them to navigate complexities of the law and procedures thereby enhancing access to justice.

Risks

Regardless of the numerous obvious benefits generative AI could offer, the Courts in Hong Kong and in overseas jurisdiction have raised concerns over the reliability of AI in recent years.

In Yu Hon Tong Thomas v Centaline Property Agency Limited [2025] HKCFI 808 (in Chinese only), the Hong Kong Court heavily criticized the use of AI in drafting by "making up severe allegations without any evidential basis".[9] The Court further challenged the addendum (as additional written submissions) to be "artificially ‘constructed’ using artificial intelligence".[10] The Court went on and noted that "the content of the whole [of the addendum] is vague and far-fetched" and a Canadian case cited was not applicable to the interpretation of the relevant Hong Kong statutory provision in question.[11]

Similarly, the Hong Kong Court expressed serious concerns over non-existent case authorities lodged by a litigant in person in Licksun Company Limited v Occupiers of Lot No. 552 in D.D. 85, New Territories, Hong Kong & Ors [2025] HKDC 1287. In particular, the Court emphasised that the lack of legal representation "provides no excuse for its lodging the Suspicious Authorities with this Court. A litigant in person is still under the duty not to mislead the Court." The Court disapproved such act and highlighted that "[r]egardless of the Plaintiff’s subjective intention, its conduct has the effect of misleading this Court", also resulting in the opponent’s solicitors to have wasted time and resources in researching cases which do not exist.[12]

The condemnation of the use of fake authorities is also found in decisions of other jurisdictions. In R. (on the application of Ayinde) v Haringey LBC and Al-Haroun v Qatar National Bank QPSC [2025] EWHC 1383 (Admin), which concerns the use of non-existent case authorities, the citations of which were generated by AI, although the English Court considered that on the available evidence, the parties concerned did not deliberately place false material before the court and did not commit the criminal offence of perverting the course of justice and contempt, the Court made a referral of the solicitors concerned to the Solicitors Regulation Authority.[13] A wasted costs order was imposed.[14]

References to cases which are not genuine are also criticised by the Australian Court[15] , the New Zealand Court[16] and the Court of British Columbia[17] .

These remarks by Courts underscore a common theme – while generative AI offers convenience and accessibility, it creates significant risks when used without prudence, profound understanding and verification. Over-reliance on generative AI without proper oversight can result in severe consequences, for instance, by lodging fabricated or baseless submissions and materials, thereby wasting judicial resources and costs, and potentially misleading the Court and leading to criminal consequences.

Guidelines and Limits

Hong Kong Guidelines

The 2024 Judiciary Guidelines[18] remain the sole piece of legal specific AI guideline in Hong Kong. Despite being at its preliminary stage, the 2024 Judiciary Guidelines largely align with the general principles in international guidelines. The 2024 Judiciary Guidelines can be rounded up into two guiding rules:

• Rule 1: No delegation of judicial functions to AI. Judicial decisions should be made independently and ensure not to allow AI to usurp or encroach upon judicial functions. [19]
• Rule 2: Check to ensure accuracy and accountability. "Be vigilant about the output generated by the AI chatbots, in particular factual accuracy, potential bias, infringement of intellectual property right, and use it at your own risk. You take responsibility for using AI and for the end product".[20]

These rules echo the concerns of the Hong Kong Courts about the limitations of AI as stated above. They also reinforce the importance of maintaining human oversight in judicial decision-making, at least until a generative AI model is developed with proven ability to protect confidential information and with a built-in checking and verification mechanism in the legal context.[21]

In addition to the above, in April 2025, the HKSAR Government commissioned the HKGAI to conduct research and formulate the ‘Hong Kong Generative Artificial Intelligence Technical and Application Guideline’ ("2025 HK Guideline") to provide guidelines on use of AI technologies in different sectors. Like the 2024 Judiciary Guidelines, the 2025 HK Guideline emphasises that "AI-generated content should not be used directly as legal documents but should serve as a reference after review by legal professionals."[22] The 2025 Guideline also highlights that "sensitive legal cases involving personal data, trade secrets and private information should not be processed using public AI services that lack security and confidentiality guarantees".[23]

International Guidelines

The Chartered Institute of Arbitrators ("Ciarb") published a ‘Guideline on the Use of AI in arbitration (2025)’[24] ("Ciarb Guideline") in March 2025, aiming to support participants of arbitration in leveraging AI technologies while mitigating some risks of use of AI. The Ciarb Guideline distinguishes itself from the existing framework in Hong Kong by offering a more well-rounded approach to disclosure, verification and party autonomy.

The Ciarb Guideline summarises the benefits (such as enhancing efficiency, improving translation and interpretation, and expediting data analysis) and risks (such as compromising confidentiality and cybersecurity and concerns about due process).[25]

It also recognises the inherent complexity and opacity of AI systems, and the challenge to comprehend the precise mechanism by which an AI tool arrives at a particular outcome, which is referred to as the "black box problem". Parties are therefore encouraged to disclose their use of AI to ensure fairness and transparency in the process, preserve the integrity of arbitration and safeguard the validity and enforceability of the award.[26] Likewise, arbitrators are also expected to disclose to parties their use of AI.

The use of AI falls within the general power of arbitrators to conduct the proceedings. Arbitrators can therefore give directions and make procedural order in relation to the same and may appoint AI experts to assist with the understanding or use of AI.[27]

In making decisions, arbitrators should respect the choice of the arbitrating parties to use AI, while independently assessing the admissibility of AI-generated content. If necessary, arbitrators should also encourage discussion between parties on the selection of AI tools. Further, similar to the principles set out in the 2024 Judiciary Guidelines, arbitrators should not relinquish their decision-making powers to AI and should always ensure independent judgment.

More recently, China International Economic and Trade Arbitration Commission ("CIETAC") issued its new Provisional Guidelines on the Use of Artificial Intelligence Technology in Arbitration (effective 18 July 2025).[28] Amongst others, it highlights three core principles of parties’ autonomy of the use of AI, the role of AI to assist with adjudication and good faith, benefits and risks of using AI, as well as guidance to the tribunal and recommended risk mitigation measures.

Proper Use of Generative AI

Like humans, computers can make mistake. It is the professional duty of legal professionals to verify the accuracy of the output of generative AI before using it in professional work, to uphold the duty owed to clients to act competently, and most importantly, to act professionally as officers of the Court. Leveraging the increasing use of AI in dispute resolution requires raising awareness of the benefits of AI but managing risks at the same time.

Users must consider the level of sensitivity of data when selecting the appropriate AI tools. They must only engage AI tools with a degree of openness that is proportional and vetted; and ensure that the AI tools meet the required security and privacy standards. Users must also bear the regulatory framework in Hong Kong in mind, such as the Personal Data (Privacy) Ordinance (Cap. 486) and the Copyright Ordinance (Cap. 528).

Finally, to secure Hong Kong’s position as the "leading centre for international legal and dispute resolution services in the Asia-Pacific region" with the proper use of lawtech[29] , it is anticipated that AI tools tailored to legal contexts will be further developed, with the appropriate safeguards to ensure that they are fit for use in legal work.