Legal AI for Arbitration in Singapore: How It's Changing Case Preparation

Legal AI for Arbitration in Singapore: How It's Changing Case Preparation

Abstract — Legal AI for arbitration Singapore practitioners use is concentrated in case preparation: research, evidence and brief summarisation, precedent finding, chronology building and first-draft submissions. Singapore's position as a leading arbitration seat makes the ethical layer unusually important, and international guidance on AI disclosure now exists. This guide covers the use cases, the tribunal considerations, and five questions before AI touches a matter.

Singapore's continued growth as a global arbitration seat, anchored by the Singapore International Arbitration Centre (SIAC), the Singapore International Commercial Court (SICC) within the General Division of the High Court, and the Singapore International Mediation Centre, has drawn attention to how legal AI for arbitration Singapore counsel deploy is actually used. International arbitration is document-heavy, multilingual and expensive, which makes it an obvious candidate for compression. It is also confidential, procedurally flexible and governed by party agreement, which makes the ethics unusually consequential.

Almost all legal AI content addresses litigation and corporate work generally. This addresses the arbitration workflow.

5 Ways Arbitration Practitioners Are Using Legal AI

1. Legal research across seats and systems. Arbitration routinely engages more than one legal system: the substantive law of the contract, the law of the seat, and the enforcement jurisdiction. Retrieval-based research handles that breadth better than any keyword approach. In Singapore the framework is the International Arbitration Act 1994 for international arbitrations, which gives effect to the UNCITRAL Model Law, and the Arbitration Act 2001 for domestic ones.

2. Summarising evidence and the document record. The largest volume saving in arbitration. Bundles running to thousands of pages, witness statements, expert reports and exhibits, compressed into summaries a team can navigate. This is the oldest and most mature application, predating generative models.

3. Precedent and authority finding. Locating awards, decisions and commentary supporting a position, and identifying analogous fact patterns. Note the structural limit: arbitral awards are largely confidential and unpublished, so the searchable body of arbitration precedent is far thinner than the case law in litigation.

4. Chronology and issue mapping. Extracting dates, events and references from the record into a chronology, and mapping the issues against the pleadings. Tedious, error-prone by hand, and well suited to extraction.

5. First-draft submissions and correspondence. Structuring memorials, procedural correspondence and factual sections. The saving is the blank page and the assembly, not the argument.

Direct answer: Arbitration practitioners use legal AI for cross-jurisdictional research, summarising large document records, finding authorities, building chronologies and issue maps, and drafting first versions of submissions. Awards are largely unpublished, which limits precedent search more than in litigation.

Time Saved: Traditional vs AI-Assisted Case Prep

No percentages here, deliberately. Published time-saving figures specific to Singapore arbitration do not exist, and a number invented for an article has a way of ending up in a costs submission.

Ethical and Tribunal Considerations

Arbitration has moved faster than most fields on AI governance, and counsel practising in Singapore should know the landscape.

International guidance exists and is being adopted. The Silicon Valley Arbitration and Mediation Center published its 2024 Guidelines on the Use of Artificial Intelligence in Arbitration, and the Chartered Institute of Arbitrators published its Guideline on the Use of AI in Arbitration on 19 March 2025, which builds on the SVAMC guidelines and includes a sample agreement and procedural order. In the region, CIETAC issued its September 2025 guidelines on AI use in arbitration. These are guidance rather than binding rules, and they apply when parties or a tribunal adopt them.

Disclosure thresholds differ between the guidelines. The CIArb guideline ties disclosure to the impact of AI use on the evidence or the outcome; the SVAMC approach leaves it to the tribunal's discretion case by case. The CIArb guideline also suggests arbitrators consult the parties before imposing disclosure requirements, and recommends refraining from AI use where the parties disagree. The practical consequence is that AI use should be raised at the first procedural conference, so expectations are set in the procedural order rather than argued about later.

Confidentiality is the sharpest Singapore-specific issue. Arbitration is confidential, and that confidentiality is a substantive obligation, not a courtesy. Uploading a confidential record to a platform whose terms permit training on inputs is a potential breach independent of any data protection question. Under the Personal Data Protection Act 2012 your organisation also stays accountable for personal data in the record once it reaches a vendor.

Fabricated authority is a professional risk, not merely an embarrassment. For advocates and solicitors, rule 5 of the Legal Profession (Professional Conduct) Rules 2015 imposes duties of honesty, competence and diligence, and the Ministry of Law's Guide for Using Generative AI in the Legal Sector, published in March 2026, sets a lawyer-in-the-loop expectation with output verified before use.

Note what does not exist. No SIAC-specific guidance on AI use was identified at the time of writing. Practitioners should check the current position with the institution administering their arbitration rather than assuming the international guidelines have been formally adopted.

Why arbitration is a harder case than litigation

Three features of arbitration make the AI question genuinely different, and counsel who reason from litigation experience get them wrong.

The precedent problem runs backwards. In litigation, AI gets better as the published corpus grows. In arbitration, the corpus is deliberately closed: awards are confidential and mostly unpublished, so the material that would make AI most useful is the material that does not exist in searchable form. A tool confidently summarising "the arbitral position" on a point should be treated with more suspicion than the same tool summarising Singapore case law, because it has far less to work from and the same fluency.

Confidentiality is substantive, not administrative. In litigation the record is largely public and the concern is client privilege. In arbitration the entire proceeding is confidential as a matter of obligation, and a disclosure to a third party platform is capable of breaching that obligation independently of any data protection analysis. The question is not only "is this secure" but "was I permitted to disclose this at all".

Procedure is party-determined. There is no equivalent of a court's practice directions applying automatically. What is permitted depends on the arbitration agreement, the institutional rules and the procedural order, which means the answer can differ from matter to matter within the same firm. That is why raising it at the first procedural conference matters more here than a general firm policy does.

The combined effect is that arbitration rewards a narrower, better-supervised use of AI than litigation does: heavy on record summarisation and chronology, cautious on anything that touches authority or the confidential record.

5 Questions Before Bringing AI Into an Arbitration Matter

1. Does the tribunal need to know, and when? Raise it at the first procedural conference. Getting AI use into the procedural order removes the risk of a dispute about it at a point when the tribunal is deciding something else.

2. Do the confidentiality obligations permit it? Check the arbitration agreement, the institutional rules and any confidentiality order. Then check the platform's terms on training and retention. Both must line up.

3. Have we agreed this with the client? For a matter of any size, the client should know which tools touch their confidential material, and the engagement terms should reflect it.

4. Who verifies, and is that time budgeted? Every authority checked against the source, every extracted date checked against the document. If verification is not resourced, the saving is illusory and the exposure is real.

5. What is our position if the other side asks? Have an answer prepared about what was used and how it was supervised. A confident, specific answer is straightforward. An improvised one is not.

Where Ask.Legal Fits Arbitration Case Preparation

Ask.Legal supports the lower-risk end of the arbitration workflow this guide describes: cross-jurisdictional research and document analysis grounded in real statutes, with a stated confidentiality position that never uses queries for model training. It is not a substitute for the tribunal disclosure discussion this guide insists on raising at the first procedural conference, and nothing in the platform changes the confidentiality obligations that govern whether a record may be uploaded anywhere at all.

For teams weighing which parts of a matter are appropriate for AI assistance more generally, the legal AI ethics in Singapore guide sets out the same verification and confidentiality standard in more depth. You can test the research and summarisation use case directly at ask.legal/en/chatbot.

Frequently Asked Questions

How is AI used in arbitration case preparation? For cross-jurisdictional research, summarising large document records, finding authorities, building chronologies and issue maps, and drafting first versions of submissions.

Do I have to disclose AI use to an arbitral tribunal? It depends on the applicable guidance and the procedural order. The CIArb guideline ties disclosure to impact on evidence or outcome; the SVAMC approach leaves it to tribunal discretion. Raise it early.

Is it safe to upload a confidential arbitration record to an AI platform? Only if the platform's terms and the confidentiality obligations both permit it. Check whether inputs are used for training, and check the arbitration agreement and any confidentiality order.

Does SIAC have guidance on AI use? No SIAC-specific guidance was identified at the time of writing. Check the current position with the administering institution rather than assuming international guidelines apply.

Can AI find arbitral precedent? Only partially. Awards are largely confidential and unpublished, so the searchable body of arbitration authority is far thinner than litigation case law.

Key Takeaways

• AI compresses research, record summarisation, chronology building and first drafts. It does not touch advocacy or strategy.

• Unpublished awards structurally limit precedent search in arbitration in a way that does not apply in litigation.

• SVAMC and CIArb guidelines on AI in arbitration exist and differ on disclosure thresholds; raise AI use at the first procedural conference.

• Confidentiality is the sharpest risk. Platform terms and the arbitration's confidentiality obligations must both permit the use.

Sources

• International Arbitration Act 1994 and Arbitration Act 2001 — Singapore Statutes Online

• Personal Data Protection Act 2012 — Singapore Statutes Online

• Ministry of Law — Guide for Using Generative AI in the Legal Sector

See how Ask.Legal supports faster arbitration case preparation

This article is general information about the law of Singapore as at 2026, not legal advice. For advice on your circumstances, consult a qualified advocate and solicitor.