AI in dispute resolution in Southeast Asia: the rules, country by country

The short version

• Courts set most of the region’s AI rules: a voluntary framework from ASEAN’s chief justices (2022, under review since July 2026), Singapore’s guide for court users (since 1 October 2024) and the Philippine Supreme Court’s framework (18 February 2026).
• Singapore allows generative AI in court documents. Users are “fully responsible” for them and declare AI use only if the court asks (Guide paras 3(3), 5(1)).
• Singapore’s High Court has made lawyers pay costs from their own pocket for citing cases that do not exist, linked to AI: S$800 in one case and S$5,000 each for two solicitors in another.
• The Vietnam International Arbitration Centre Rules 2026 let the tribunal discuss AI tools with the parties (art. 16(6)). The rules we read from Malaysia, Indonesia and the Philippines say nothing on AI.
• ASEAN’s AI governance guides are voluntary and do not mention courts or disputes. The ASEAN Guidelines on Online Dispute Resolution (ODR) see fully automated decisions as a long-term option for simple consumer cases (para 49).
• Vietnam’s 35-article Law on Artificial Intelligence took effect on 1 March 2026, according to the Ministry of Science and Technology. We found nothing official on how it applies to courts or arbitration.^{1,3,5,9,12,16,21,22,25,26,28,31,34,35,36,37}

How is AI used in dispute resolution in Southeast Asia?

In the official sources we found, AI in dispute resolution in Southeast Asia mostly means courts’ own support tools: translation, summaries, transcription and sentencing references. One Singapore mediation centre launched a drafting assistant in 2024. Arbitration rules are largely silent.^{8,16,19,31,34,36,37,38}

This guide covers South East Asia’s ASEAN member states. It uses only the texts of courts, institutions, governments and intergovernmental bodies, checked between 23 and 26 October 2026 (how we research).

Where we could not open a body’s own site, we say so rather than rely on reports. That applies to the Singapore International Arbitration Centre (SIAC), the Thailand Arbitration Center (THAC), Indonesia’s Supreme Court and Vietnam’s Supreme People’s Court portal.

Court use started from a low base. In a survey for the AI framework of the Council of ASEAN Chief Justices (CACJ), every member state at the time except Malaysia and Singapore had not adopted AI in its courts. Respondents blamed weak technical infrastructure, legal and policy limits and scarce resources (printed p. 5). The survey is undated.²

Singapore’s Small Claims Tribunals have heard over 10,000 cases a year on average in recent years. Parties have had generative AI translations of their Notice of Consultation since December 2024, and of all documents they file since March 2025 (September 2025 release, para 7). A case-summary tool has served the tribunals’ magistrates and registrars since 10 September 2025 and self-represented parties since 2 December 2025 (December 2025 release, paras 2, 8, 9). It gives factual summaries without case-specific legal advice (September 2025 release, para 5).^7,8

Courts in Sabah and Sarawak began using AI to recommend sentences for drug possession in February 2020 (CACJ framework, printed p. 5). On 22 July 2021 Malaysia’s judiciary announced an AI sentencing guideline for the internal use of Sessions Court judges and magistrates in Peninsular Malaysia, piloted from 23 July 2021 as “only a guide and quick reference” (our translation).^2,19

The Philippine Supreme Court piloted voice-to-text transcription of stenographic notes (A.M. No. 25-11-28-SC, recitals, p. 2). In October 2026 the Court reported one of its Justices as saying that a transcription tool used by stenographers nationwide was the only AI tool the Court had allowed, and that a stenographer reviews the text before it becomes the official transcript.^16,18

In a November 2025 story, UNESCO quoted a Thai Supreme Court judge as saying that Thailand’s justice system, like Brunei’s, makes no formal use of AI. She said it is exploring a tool to explain procedures, estimate filing costs and help draft simple claims, without assessing merits or predicting outcomes.³⁰

On 28 August 2024 the Singapore International Mediation Centre (SIMC) launched an AI assistant that builds chronologies, maps who is involved, summarises disputes and can help draft an initial settlement agreement “for counsel to revise and finalise”. The Singapore Mediation Centre published a prompt guide for mediators in August 2026. Our dispute resolution guide covers how AI is used in mediation beyond the region.^38,39

For arbitration, see the [institutions table](#institutions) below and our guide to how AI is used in arbitration. We did not survey the region’s ombudsman schemes for this version. Our other regional guides cover Australia and Europe.

What rules have courts in Southeast Asia set on AI?

Of the ASEAN judiciaries whose own texts we could read, two have general rules on AI use in court: Singapore’s guide for court users, applied since 1 October 2024, and the Philippine Supreme Court’s framework of 18 February 2026. Both make the person using AI responsible for its output. Malaysia’s 2021 guideline covers only an AI sentencing tool for judges.^5,16,19

Singapore’s Guide on the Use of Generative Artificial Intelligence Tools by Court Users covers the Supreme Court, the State Courts (including the Small Claims Tribunals) and the Family Justice Courts (para 1). It has applied since 1 October 2024 under Registrar’s Circular No. 1 of 2024 and the other courts’ circulars ([2025] SGHCR 33 at [37]).^5,6,9

The Guide does not prohibit generative AI, and users need not declare its use unless the court asks (paras 3(1), 3(3)). Users are “fully responsible” for their documents. They should not use AI to generate evidence, though a first-cut draft of an affidavit is allowed, and should check every authority against a source known to be accurate, never with another AI tool (paras 5(1), 5(3)). Breaches can bring personal costs orders, which make lawyers pay from their own pocket (para 5(9)).⁵

The Ministry of Law’s Guide for Using Generative AI in the Legal Sector (6 March 2026) is non-binding. It is meant for anyone handling legal work in Singapore and complements the courts’ guide (paras 2 to 4). Our law and legal tech guides cover what courts elsewhere expect and what professional rules require of lawyers using AI.¹⁴

The Philippine Supreme Court, sitting as a full court (En Banc), adopted A.M. No. 25-11-28-SC on 18 February 2026 and announced it on 19 March. It is a framework for “human-centered augmented intelligence” covering judges, court staff, court users (including members of the Bar) and vendors (Parts I.B, I.C.9).^16,17

Judges and court staff who use AI must disclose, at minimum, the tool and version, the purpose, the extent of use and the human oversight, and state that they kept the output, complied with the framework and bear ultimate responsibility. Disclosure is mandatory for court-issued documents in adjudicatory work, under guidelines the Court will issue (Part II.3, PDF pp. 9, 10).¹⁶

Every user is “personally responsible” for AI output, and AI must never be “the sole, primary or determinative basis of any adjudicatory outcome” (Part II.4, PDF pp. 10, 11). Confidential or privileged information may go into an AI tool only with the Court’s express authority (Part II.6). No AI tool may be used unless the full Court authorizes it, and the Court decides case by case how far those covered may use AI (Parts III.1.A, III.3, PDF pp. 14, 19).¹⁶

We did not check Cambodia, Lao PDR and Myanmar beyond the CACJ survey, or Timor-Leste, admitted to ASEAN as its 11th member state on 26 October 2025, which had joined the CACJ by July 2026.^3,4

What have ASEAN and its chief justices said about AI in courts?

The Council of ASEAN Chief Justices (CACJ) approved a voluntary Governance Framework on the Use of Artificial Intelligence for the ASEAN Judiciaries on 4 November 2022. It keeps AI as a support tool under human control and has been under review since July 2026. The ASEAN Guide on AI Governance and Ethics (2024) and its generative AI expansion (January 2025) are also voluntary (Introduction; Conclusion) and do not mention courts or disputes.^1,2,3,25,26

The Kuala Lumpur Declaration recommended the framework for adoption under each member state’s own laws, “with such modifications as each jurisdiction deems fit”. Member states “may voluntarily opt to use this framework” (Part A, printed p. 7). Its five principles are inclusivity and sustainability, fairness and human oversight, transparency and explainability, robustness and safety, and accountability and auditability (Part B, printed p. 22).^1,2

The framework recommends that the justice system’s use of AI “should primarily be in the form of support systems and augmentation tools and not completely supplant human judgment” (printed p. 23). Oversight can be a human in the loop, able to step in on every decision, a human on the loop or a human in command. In each case, users “must be able to deviate from the outcome of the algorithm without difficulty” (printed p. 29). The text does not mention generative AI.²

On 22 July 2026 the chief justices agreed that a working group would review and revise the framework and report at their next meeting (Bangkok Declaration, para 15). We found no revised text.³

UNESCO published Guidelines for the use of AI systems in courts and tribunals in 2025. With the UN Development Programme and the Thailand Institute of Justice, it ran a workshop in Bangkok from 12 to 14 November 2025 for 27 judges from 13 Asia-Pacific countries, the first such workshop in Southeast Asia, it says. The judges weighed AI’s promise for access to justice against risks such as bias and lost public trust.^29,30

Do arbitral and mediation centres in Southeast Asia have AI rules?

Few do. In arbitration in Southeast Asia, the Vietnam International Arbitration Centre Rules 2026 are the only rules we could read that mention AI. The Asian International Arbitration Centre (AIAC) Arbitration and i-Arbitration Rules 2026, the BANI Arbitration Rules 2025 and the 2021 rules of the Philippine Dispute Resolution Center (PDRCI) do not. In mediation, two Singapore centres have an AI tool and a prompt guide rather than rules.^{31,34,35,36,37,38,39}

The Vietnam International Arbitration Centre Rules have been in force since 1 July 2026. Under art. 16(6), the tribunal “may discuss with the parties the use of information technology tools in the arbitral proceedings, including tools supported by or incorporating artificial intelligence”. It can do so at a case management conference, which it may call early on (art. 16(2)), or at any other appropriate stage.³¹

The wording is close to South Korea’s KCAB International 2026 Rules, under which the tribunal “is encouraged to discuss” AI tools (art. 16.5). Neither limits AI use or restates non-delegation, the principle that an arbitrator cannot hand the decision to anyone else. For a fuller institutional text in Asia, see China’s CIETAC trial guidelines of 18 July 2025, which sit outside its rules.^31,32,33

For SIAC, see our tracker’s entry. SIAC’s rules and practice are covered in our guide to arbitration in Singapore. Our arbitration guide sets out the AI guidelines arbitrators use worldwide.

How do ASEAN’s online dispute resolution guidelines and Singapore’s small claims tribunals treat AI?

ASEAN’s consumer guidance on online dispute resolution (ODR) treats AI as a later step, from screening claims to, in the long term, fully automated decisions. Singapore’s Small Claims Tribunals use generative AI to translate and summarise, and a Tribunal Magistrate hears the case.^8,28

The ASEAN Committee on Consumer Protection launched the ASEAN Guidelines on Online Dispute Resolution on 8 April 2022, when only Indonesia, the Philippines and Thailand had a national ODR system. On automation they say:^27,28

• Algorithms and AI “are also employed” in ODR systems (para 16).
• A fully automated system should not be the aim from the start (para 43).
• Where technology allows, algorithms could screen whether claims are valid (para 46).
• Fully technology-enabled decisions on compensation, “involving AI and predictive justice”, need careful design to be fair and unbiased. They are a long-term option, most feasible for straightforward cases with clear rules and redress criteria (para 49).²⁸

Singapore’s Small Claims Tribunals hear claims of up to S$20,000, or S$30,000 if both parties agree in writing (para 2). For ODR standards and online tribunals outside the region, see our dispute resolution guide.⁸

What does the law say about AI and automated decisions in Southeast Asia?

Of the general AI texts we checked, from Vietnam, Malaysia, Singapore and ASEAN, only Vietnam’s is a law. The others are guidelines or frameworks that do not address AI use by courts or arbitral bodies, and the Vietnamese ministry’s pages do not say whether the law does.^{15,20,21,22,25,26}

On 15 December 2025 Vietnam’s Ministry of Science and Technology reported that the National Assembly had approved a 35-article Law on Artificial Intelligence, which requires human oversight “in important decision-making processes”. On 5 March 2026 it reported the law in force since 1 March. It has since reported an AI ethics circular (Circular No. 05/2026/TT-BKHCN) requiring human oversight “in all decisions and actions of AI systems”, scaled to their impact, and a Prime Minister’s decision listing high-risk AI systems (Decision No. 33), which took effect on 15 August 2026. We have not read these texts themselves.^21,22,23,24

Malaysia’s National Guidelines on AI Governance and Ethics (September 2024) set seven principles for voluntary adoption (Executive Summary; s 1.3). They say consumers must be able to object to, and get an explanation of, decisions an algorithm makes with their data, and to deal with a human instead of AI (Part A, s 2.3, printed p. 40).²⁰

Singapore’s Model AI Governance Framework for Generative AI (30 May 2024) proposes nine dimensions for a trusted AI ecosystem (Executive Summary), and the Ministry of Law’s guide builds on it (para 2). Our law guide covers how AI itself is regulated.^14,15

Every court text we found keeps the decision with a human. Only the ASEAN ODR Guidelines contemplate fully automated decisions, as a long-term option for simple consumer cases (para 49). We have not checked national data protection laws, so we make no claim about rights over automated decisions under them.^2,16,19,28

Have Southeast Asian courts sanctioned AI misuse?

Yes, in Singapore. Four High Court judgments from September 2025 to March 2026 dealt with fictitious authorities, fake cases cited as if real, where generative AI was used or suspected. This kind of invented output is called a hallucination. We found no official decision on AI misuse elsewhere in the region in the sources we could open.^9,10,11,12

• Tajudin bin Gulam Rasul and another v Suriaya bte Haja Mohideen [2025] SGHCR 33 (29 September 2025), decided by an assistant registrar: counsel cited a case a generative AI tool had invented (para 5) and was ordered to pay S$800 in personal costs (para 99). Lawyers may not delegate research and drafting to generative AI “without independently verifying the AI-generated output” (para 101).
• Landscape Engineering Pte Ltd v Dot Safety Solutions Pte Ltd and another [2025] SGHC 214 (30 October 2025): a company officer relied on generative AI and cited “no fewer than six fictitious cases” (para 40). The court warned of costs against him personally if it happened again (para 50).
• Forbes Monaco APAC v Kawajiri Seiji [2026] SGHC 16 (21 January 2026): a self-represented person relied heavily on generative AI, failed to disclose it and caused at least nine fictitious authorities to be cited (para 78). He was ordered to pay costs personally (para 2).
• Tan Hai Peng Micheal and another v Tan Cheong Joo and another and other matters [2026] SGHC 49 (6 March 2026): whether AI produced the two fictitious authorities was “immaterial” (para 32). The S$1,500 the two solicitors had offered was “plainly inadequate”, and each was ordered to pay S$5,000 personally (paras 63, 66).^9,10,11,12

In grounds of 4 May 2026, an assistant registrar also ordered counsel to pay S$900 personally for miscited cases, including a “wholly fictitious quotation”, whether or not AI was used ([2026] SGHCR 13, paras 91, 98). All are court cases; we found no published arbitral or mediation decision in the region on AI misuse. For fake citations elsewhere, see our law guide.¹³

Which Southeast Asian states have joined the Singapore Convention on Mediation?

Singapore is the only ASEAN member state that is a party to the Singapore Convention on Mediation, the UN treaty on enforcing international mediated settlements. It ratified on 25 February 2020. Brunei, Lao PDR, Malaysia, the Philippines and Timor-Leste signed on 7 August 2019, and the status page of the UN Commission on International Trade Law (UNCITRAL) lists no ratification for them.⁴⁰

Cambodia, Indonesia, Myanmar, Thailand and Vietnam are not on the list. The Convention has been in force since 12 September 2020, with 25 parties according to the UN Treaty Collection as at 23 October 2026. It covers international commercial settlements, not consumer ones (art. 1(2)(a)), and does not address AI. Our dispute resolution guide asks whether AI can act as a mediator.^41,42

What should parties, counsel and institutions in the region do about AI now?

Follow the court’s or centre’s own AI text where there is one, and agree AI terms early where the rules say nothing. Each step below rests on a text cited above:

• Before Singapore’s courts, check that every case, statute or quote from AI exists and says what you cite it for, using an accurate source, never another AI tool (Guide para 5(3)). Be ready to show which passages used AI and how you verified them (para 5(4)). Lawyers risk personal costs ([2025] SGHCR 33 at [99]; [2026] SGHC 49 at [66]).
• Before Philippine courts, you are “personally responsible” for AI output (Part II.4). Keep confidential or privileged information out of AI tools unless the Court authorizes it (Part II.6).
• Under the Vietnam International Arbitration Centre Rules 2026, ask for a case management conference (art. 16(2)) or otherwise invite the tribunal to discuss AI tools (art. 16(6)).
• Where rules say nothing on AI, as with the AIAC, BANI and PDRCI, agree AI terms early, for example in Procedural Order No. 1, the first order setting a case’s ground rules. Our arbitration guide lists practical steps for tribunals.
• Institutions building or buying AI tools can borrow from texts written for courts and consumer ODR: keep users able to deviate from the output (CACJ, printed p. 29), avoid full automation at the start (ASEAN ODR Guidelines para 43), and record the tool, version, purpose, extent of use and oversight (Philippine framework, Part II.3).
• Mediators can treat AI as “an aid, not a substitute for professional judgment” and be transparent with the parties about using it (Singapore Mediation Centre guide, rules 1 and 6).
• For cross-border settlements, check whether the state where you may need to enforce is a party to the Singapore Convention; in ASEAN, only Singapore is.^{2,5,9,12,16,28,31,34,35,36,37,39,40,42}

What is still open on AI in dispute resolution in Southeast Asia?

The biggest open questions are how Vietnam’s AI Law applies to courts and arbitration, and what the chief justices’ review of their framework will change. Others:^3,22

• How far the Philippine framework governs lawyers’ own tools. It covers court users’ submissions (Part I.C.6), but its disclosure duty is written for judges and court staff (Part II.3), and we found no disclosure guidelines issued under it.
• Which judiciaries have adopted the CACJ framework. The only court text we found that cites it is the Philippine resolution, which adopts the Supreme Court’s own framework (recitals, p. 2).
• Whether the AIAC, BANI, PDRCI and other arbitral centres will address AI in their rules, and what SIAC and THAC, whose sites we could not open, have said.
• How often Singapore’s courts will expect disclosure. The Guide requires a declaration only if the court asks (para 3(3)), but in [2026] SGHC 16 the court noted a failure to disclose heavy AI use (para 78).
• What safeguards would apply to the automated consumer decisions the ASEAN ODR Guidelines contemplate (para 49).
• Whether a settlement reached on flawed AI advice could be refused enforcement under the Singapore Convention, one of the open questions in our dispute resolution guide.^{1,5,11,16,28,34,36,37,42}