CIVIL LAW AND UAE – RIGHT TO EXPLANATION IN AUTOMATED JUDGMENTS

1. Introduction

The concept of a right to explanation in automated judgments concerns the ability of a litigant to understand how an automated or AI-assisted system contributed to a judicial outcome.

The central question is:

If artificial intelligence materially assists in producing a civil judgment, can the affected party understand the facts, evidence, legal rules and reasoning that led to the result?

This question is becoming increasingly important because courts can use technology for:

document classification;

evidence analysis;

legal research;

case management;

transcription;

translation;

calculation;

scheduling;

decision-tree systems;

digital dispute resolution;

artificial-intelligence-assisted forms; and

other judicial-support functions.

The UAE has already developed sophisticated digital-court infrastructure. The DIFC Digital Economy Court, for example, expressly covers disputes involving artificial intelligence and permits AI-driven smart forms for the conduct and disposal of claims.

However, automation of judicial processes does not necessarily mean automation of judicial responsibility.

The basic legal model should therefore be:

AI assistance → human verification → judicial reasoning → reasoned judgment → appeal/review

rather than:

AI input → unexplained algorithmic output → automatic judgment.

2. Meaning of the Right to Explanation

A right to explanation can be understood as a procedural entitlement to receive sufficient information about the basis of a legally significant automated or AI-assisted decision to enable the affected person to:

understand the outcome;

identify the decisive factual findings;

understand the legal rules applied;

identify important evidence relied upon;

understand material objections and defences;

identify the role played by automated technology;

challenge errors;

seek appellate review; and

correct inaccurate information where appropriate.

It does not necessarily require disclosure of every line of computer code or every proprietary algorithm.

The essential requirement is meaningful legal intelligibility.

3. UAE Position

There is currently no single UAE civil-law provision expressly stating:

"Every person has a statutory right to an explanation of an AI-generated judgment."

Instead, the principle is constructed from existing legal requirements.

The most important foundation is the UAE Civil Procedure Code.

4. Federal Decree-Law No. 42 of 2022

Federal Decree-Law No. 42 of 2022 promulgates the UAE Civil Procedure Code.

Article 129 requires judgments to state their underlying grounds.

Article 130 further requires a judgment to contain, among other things:

the relevant facts;

the relief sought;

a summary of the parties' defences;

the reasons for the judgment; and

the operative part.

Failure to satisfy specified requirements concerning the grounds and contents of the judgment can affect its validity.

This is extremely important for automated adjudication.

A judgment cannot ordinarily satisfy the legal requirement of reasoning merely by saying:

"The automated system determined that the claimant is liable."

The judgment should explain the legally significant reasoning connecting:

Facts → Evidence → Law → Analysis → Conclusion

5. Article 129 and Automated Judgments

Article 129 is particularly significant.

The requirement that judgments indicate their underlying grounds establishes a structural barrier against completely opaque adjudication.

An AI system may produce:

"Probability of liability: 87%."

But that is not, by itself, a legal reason.

The judge must still establish:

what facts were accepted;

why those facts were established;

which evidence was accepted;

which evidence was rejected;

what law applies;

how the law applies to the facts;

why competing arguments were rejected; and

why the particular remedy follows.

Thus:

Algorithmic output ≠ judicial reasoning.

6. Article 130 and Explainability

Article 130 is even more directly relevant.

The judgment must contain a factual account, the claims and defences, and the grounds and operative part of the judgment.

This creates several layers of explanation.

Layer 1 – Factual explanation

What happened?

Layer 2 – Evidentiary explanation

What evidence established the facts?

Layer 3 – Legal explanation

Which legal rules apply?

Layer 4 – Analytical explanation

Why do those rules produce the result?

Layer 5 – Remedial explanation

Why is the particular remedy appropriate?

An AI-assisted judgment should preserve this structure.

7. Why Explanation Matters More When AI Is Used

Human judges can ordinarily be questioned about the reasoning contained in their judgments and are subject to established procedural and appellate mechanisms.

AI systems create additional risks:

opaque algorithms;

incomplete datasets;

incorrect classifications;

hidden assumptions;

statistical bias;

automation errors;

hallucinated information;

incorrect legal authorities;

data-quality problems;

excessive confidence;

model drift.

Consequently, the more important the role of AI in producing an outcome, the more important it becomes that the final judicial reasoning remains independently intelligible.

8. DIFC Courts and AI

The DIFC Courts issued Practical Guidance Note No. 2 of 2023 concerning the use of large language models and generative AI in proceedings.

The guidance emphasises:

transparency;

accuracy;

reliability;

verification;

disclosure of AI use;

awareness of limitations;

potential bias;

confidentiality; and

avoidance of excessive reliance on AI.

It also expressly emphasises that AI should assist rather than replace the human decision-making integral to legal proceedings.

This provides an important institutional indication of how AI should be treated in the UAE's specialised commercial-court environment.

9. DIFC Digital Economy Court

The DIFC Courts' Part 58 provides a Digital Economy Court framework.

Its jurisdiction includes disputes involving:

artificial intelligence;

digital assets;

blockchain;

substantial databases;

digital data;

fintech; and

other digital-economy matters.

The Rules also contemplate smart forms and AI-driven forms, including decision-tree software that can obtain information necessary for the conduct and disposal of claims.

This demonstrates that the UAE legal system is willing to integrate AI and automation into civil justice.

But procedural digitalisation does not eliminate the obligation to provide a legally intelligible judgment.

10. Automated Judgment vs AI-Assisted Judgment

This distinction is essential.

A. AI-Assisted Judgment

A human judge remains the legal decision-maker.

AI may assist with:

research;

document review;

summarisation;

translation;

calculations;

classification.

The judge independently determines the dispute.

B. Automated Judgment

An automated system itself materially determines the outcome according to predetermined or machine-learning processes.

This raises substantially greater questions concerning:

legal authority;

judicial accountability;

procedural fairness;

explanation;

human oversight;

appeal;

error correction.

The UAE's present framework provides considerably stronger support for AI assistance under human judicial responsibility than for completely autonomous judicial decision-making.

11. The Human Judicial Decision-Maker

A central principle is:

Technology may assist judicial decision-making, but legal responsibility for the judgment must remain attributable to the competent judicial authority.

A computer can:

calculate damages;

identify similar cases;

organise evidence;

search documents;

detect patterns.

But the legal system still requires determination of:

admissibility;

credibility;

legal interpretation;

materiality;

causation;

liability;

defences;

remedy.

These are legal judgments.

12. Right to Explanation and Right of Defence

The right of explanation is closely connected with the right of defence.

A litigant cannot meaningfully challenge a decision if the decision merely states:

"The AI system found against the defendant."

The litigant needs to know:

which evidence was decisive;

which defence was rejected;

why it was rejected;

what legal rule was applied;

whether AI-generated analysis materially influenced the result.

Without such information, the right to challenge may become theoretical rather than effective.

13. Right to Explanation and Appeal

Reasoned judgments are also necessary for appellate review.

An appellate court needs to know:

what the lower court decided;

what evidence it relied upon;

what legal test it applied;

how it reached its conclusion.

If a judgment contains only an unexplained algorithmic result, the appellate court may have difficulty determining whether:

evidence was ignored;

the wrong law was applied;

a material defence was overlooked;

the algorithm contained erroneous assumptions;

the judge independently evaluated the result.

Therefore:

Explainability → effective appeal → judicial accountability.

14. Case Law

Because fully autonomous AI judgments are still an emerging issue, UAE reported jurisprudence has generally dealt with the underlying principles rather than a mature doctrine expressly called a "right to explanation in automated judgments."

The following cases are therefore particularly useful.

15. Case 1 – Oheo Bank v Parker [2025] DIFC CA 006

Principle: Adequate reasons and effective appellate review

This 2026 DIFC Court of Appeal judgment is particularly important.

The Court emphasised that adequate reasons are an essential condition for the satisfactory operation of appellate review.

The reasons must enable the appellate court to determine whether the judge:

addressed determinative issues;

considered relevant matters;

reasoned logically and rationally.

The Court described the importance of using the building blocks of the reasoned judicial process.

It criticised a judgment that consisted essentially of conclusions without explaining the reasoning leading to those conclusions.

Relevance to AI

This is highly significant for automated or AI-assisted judgments.

An AI-assisted judgment cannot simply state:

"The model produced Result X."

The judgment should expose the legally relevant reasoning behind Result X.

Principle

AI output cannot substitute for legally sufficient reasons.

16. Case 2 – UAE Civil Cassation No. 647 of 2021

Principle: Material evidence and defences must be considered

The UAE Court of Cassation has emphasised that a judgment should contain sufficient elements demonstrating that the court properly understood and examined the material facts and evidence.

A material defence capable of affecting the outcome cannot simply be ignored.

AI relevance

Imagine an AI system reviews 100,000 documents and concludes:

"There is no evidence supporting the defendant."

The defendant then identifies five documents that directly address the issue.

The court should not mechanically accept the algorithmic classification.

It must determine:

whether the documents were processed;

whether they were relevant;

whether they affect the outcome;

why the defence succeeds or fails.

Principle

Automated evidence processing does not remove the judicial obligation to consider material evidence and defences.

17. Case 3 – UAE Commercial Cassation No. 215 of 2020

Principle: Expert conclusions cannot replace judicial reasoning

The Court considered the role of expert evidence and the need for proper judicial assessment of technical conclusions.

The court may rely upon an expert where the expert's work provides a proper factual and technical foundation.

However, the expert does not become the legal decision-maker.

AI relevance

An AI system may perform a function similar to highly sophisticated technical evidence.

For example:

"The algorithm calculates that the contractor caused 73% of the project delay."

That does not automatically establish legal liability.

The court still needs to determine:

contractual responsibility;

causation;

contractual exclusions;

extensions of time;

force majeure;

mitigation;

applicable law.

Principle

Technical analysis does not equal legal adjudication.

18. Case 4 – UAE Commercial Cassation No. 767 of 2021

Principle: Experts cannot determine legal questions

The Court distinguished between:

technical/factual matters, which can be addressed by experts,

and

legal questions, which remain for the court.

Application to AI

An AI system can potentially determine:

accounting discrepancies;

numerical calculations;

document similarities;

technical defects;

statistical patterns.

But it should not independently determine:

whether a contract was breached;

whether a party acted unlawfully;

whether a limitation defence applies;

whether public policy prevents enforcement;

what legal remedy follows.

Principle

AI may provide technical assistance without acquiring judicial authority.

19. Case 5 – UAE Commercial Cassation No. 872 of 2023

Principle: Clear and sufficient judicial reasoning

The Court reiterated the importance of clear and sufficient reasons demonstrating that the court properly considered the dispute and evidence.

AI relevance

An AI-assisted judgment should not rely on unexplained statements such as:

"The algorithm determined that the evidence is unreliable."

The judgment should instead identify:

the evidence;

the reason for treating it as unreliable;

the relevant legal test;

the analysis supporting the conclusion.

Principle

Explainability is part of judicial accountability.

20. Case 6 – UAE Commercial Cassation Nos. 1012 and 1023 of 2022

Principle: Technical analysis does not replace judicial determination

These decisions reinforce the distinction between expert/technical analysis and the ultimate legal responsibility of the court.

AI relevance

An AI system may calculate:

"Estimated contractual loss = AED 2 million."

But the court must determine whether:

the loss is legally recoverable;

causation is established;

the loss is too remote;

contractual limitations apply;

mitigation was required;

the claimant proved the amount.

Principle

Algorithmic calculation is evidence or assistance, not necessarily the final legal conclusion.

21. Case 7 – Arabyads Holding Limited v Gulrez Alam Marghoob Alam [2025] ADGMCFI 0032

This ADGM case is highly relevant to AI-generated legal material.

The Court imposed substantial wasted costs after legal submissions relied on authorities that did not exist and displayed characteristics associated with AI-generated hallucinations.

The central lesson was that lawyers cannot avoid professional responsibility by relying on AI.

AI relevance

If AI-generated material is used in litigation:

AI generation → human verification → professional responsibility

must remain the chain.

Broader principle

If lawyers remain responsible for verifying AI-generated material, a judicial decision-maker likewise cannot simply outsource legal responsibility to an AI system.

22. Case 8 – Johnson Arabia LLC v BIC Contracting LLC [2020] DIFC CFI 075

This case demonstrates the importance of identifying the actual issues that require judicial determination.

The Court analysed lists of issues and distinguished genuinely determinative questions from matters that merely paraphrased broad aspects of the dispute.

AI relevance

AI systems can process huge amounts of information, but volume does not determine legal relevance.

An AI-assisted judicial process must distinguish:

data → relevant evidence → material issue → legal conclusion.

Principle

More information does not necessarily mean better legal reasoning.

23. Case Law Summary

24. What Should an AI-Assisted UAE Judgment Explain?

A robust judgment should ideally disclose enough information to establish the following chain:

1. AI involvement

Was AI used?

2. Purpose

Was AI used for:

translation;

research;

document review;

evidence classification;

calculations;

decision support?

3. Human responsibility

Who made the final legal determination?

4. Evidence

What evidence was relied upon?

5. Legal rules

What legislation or legal principles were applied?

6. Material objections

What important arguments were raised?

7. Analysis

Why were those arguments accepted or rejected?

8. Conclusion

How did the reasoning produce the result?

25. What Does Not Necessarily Need to Be Disclosed?

A right to explanation should not automatically mean disclosure of:

source code;

proprietary algorithms;

security-sensitive information;

confidential vendor information;

irrelevant technical details;

internal system architecture.

The objective should be legal explainability, not unlimited technological disclosure.

For example:

"The AI model used a proprietary neural network with 175 billion parameters."

may tell a litigant very little about the legal reasoning.

By contrast:

"The system classified these documents as relevant, but the Court independently reviewed the disputed documents and relied upon documents A, B and C because they established X."

is much more useful.

26. The Concept of Meaningful Explanation

A meaningful explanation should answer four questions:

Question 1

What happened?

Question 2

What evidence proves it?

Question 3

What law applies?

Question 4

Why does the law produce this result?

If these questions can be answered, the decision is substantially more transparent.

27. Black-Box Problem

A black-box AI system may produce:

Input → Output

without making the intermediate reasoning understandable.

For civil adjudication, this creates several problems.

Problem 1 – Error

Incorrect data may produce an incorrect result.

Problem 2 – Bias

Historical data may reproduce systematic patterns.

Problem 3 – Missing evidence

The system may fail to identify relevant documents.

Problem 4 – Legal misunderstanding

The AI may misunderstand a legal concept.

Problem 5 – Appeal

An appellate court may be unable to reconstruct how the outcome was reached.

Problem 6 – Accountability

It may become unclear who is legally responsible for the decision.

28. Automated Judgment and Natural Justice

Although terminology differs between UAE legal systems, the fundamental procedural concern is fairness.

A litigant should have meaningful opportunity to:

present evidence;

make submissions;

challenge opposing evidence;

respond to adverse material;

challenge errors;

seek review.

A completely opaque automated determination can threaten these procedural guarantees.

29. Right to Explanation and Evidence Law

Federal Decree-Law No. 35 of 2022 promulgates the UAE Law of Evidence in Civil and Commercial Transactions.

The modern evidence framework recognises electronic forms of evidence.

This is important because AI-assisted courts will increasingly deal with:

electronic documents;

electronic records;

digital communications;

databases;

electronic signatures;

digital transactions.

But electronic evidence still needs evaluation.

The fact that a computer generated information does not automatically establish:

authenticity;

accuracy;

completeness;

legal relevance;

reliability.

Therefore:

Electronic evidence → authentication → evaluation → legal reasoning

remains necessary.

30. Automated Evidence Assessment

Suppose an AI system analyses bank transactions and concludes:

"The transaction is fraudulent."

The court should separate:

Technical question

What transaction occurred?

Analytical question

What pattern does the data show?

Evidentiary question

How reliable is the analysis?

Legal question

Does the conduct satisfy the elements of fraud under applicable UAE law?

Remedial question

What remedy follows?

The AI may assist with the first three.

The final legal questions remain matters for the legally authorised decision-maker.

31. Human-in-the-Loop Model

The safest conceptual structure for judicial AI is:

AI analysis

↓

Human judicial review

↓

Challenge by parties

↓

Judicial assessment

↓

Reasoned judgment

This is a human-in-the-loop model.

It preserves:

accountability;

explainability;

procedural fairness;

correction;

appellate review.

32. Human-on-the-Loop Model

Another model is:

AI operates automatically

↓

Human monitors

↓

Human intervenes where necessary

This may be suitable for lower-risk procedural functions such as:

scheduling;

document organisation;

administrative classification.

It becomes more problematic when the AI determines substantive legal rights.

33. Human-out-of-the-Loop Model

Here:

AI → decision

without meaningful human intervention.

For judicial determination of substantive civil rights, this creates much greater concerns regarding:

statutory judicial authority;

accountability;

reasoning;

procedural fairness;

appeal;

responsibility.

The present UAE framework does not establish a general doctrine authorising fully autonomous AI courts to replace legally constituted judges.

34. CBUAE Developments

The UAE Central Bank's AI-related regulatory framework provides an important example outside ordinary court adjudication.

Its guidance requires meaningful human oversight for AI and ML systems in relevant financial-sector contexts.

It also provides that consumers should be able to request human review or explanation of AI-generated decisions, and should have channels to challenge decisions and correct inaccurate data.

Although this is a financial-regulatory context rather than a general judicial right, it demonstrates that human review and explanation are recognised as important safeguards in UAE AI governance.

35. Right to Explanation and Data Protection

AI systems may process substantial quantities of personal information.

Consequently, automated judicial or dispute-resolution systems may raise issues concerning:

accuracy;

purpose limitation;

data security;

access;

correction;

confidentiality;

lawful processing.

The interaction between data governance and civil procedure becomes especially important when an algorithmic decision is based on personal data.

36. Right to Explanation and Confidential Algorithms

There is a tension between:

Transparency

and

protection of proprietary technology.

A company operating an AI system may argue that revealing the model would expose:

trade secrets;

security vulnerabilities;

intellectual property.

The solution need not be complete disclosure.

The more appropriate approach may be:

Technical confidentiality + judicial inspection + legally sufficient explanation to the parties.

37. Judicial Explainability vs Algorithmic Explainability

These should not be confused.

Algorithmic explainability

Explains how the software produced a result.

Judicial explainability

Explains why the legally authorised decision-maker reached the legal conclusion.

A technically explainable algorithm can still produce an inadequately reasoned judgment.

Therefore:

Algorithm explanation ≠ judicial explanation.

38. Standard for a Reasoned Automated Judgment

A useful model is:

A – Attribution

Who made the decision?

B – Basis

What evidence was relied upon?

C – Rules

What legal provisions were applied?

D – Analysis

How were the facts connected to the law?

E – Defences

What material arguments were considered?

F – Result

What legal conclusion followed?

G – Remedy

Why was the particular remedy selected?

This can be called the:

A-B-C-D-E-F-G model of judicial explainability.

39. Example

Suppose an AI system evaluates a construction dispute.

AI output

"Contractor liable for AED 5 million."

That is insufficient as a judicial explanation.

A reasoned judgment should instead explain:

the contractual obligations;

the alleged delay;

project records;

expert evidence;

relevant correspondence;

causation;

extension-of-time arguments;

mitigation;

calculation of loss;

applicable legal provisions;

why the contractor's defence succeeds or fails;

why AED 5 million is legally recoverable.

AI may assist in processing the documents, but the judgment must remain legally reasoned.

40. Remedies for Defective Explanation

Where a judgment is insufficiently reasoned, possible legal consequences may include:

appeal;

cassation;

remittal;

correction or interpretation where legally appropriate;

setting aside where statutory requirements are violated;

judicial reconsideration.

The precise remedy depends on the court, jurisdiction and nature of the defect.

The important principle is:

An unexplained result should not become immune merely because a computer generated it.

41. Relationship with Judicial Independence

AI should not undermine judicial independence.

The judge must retain authority to:

reject AI recommendations;

request additional evidence;

order expert examination;

hear parties;

evaluate conflicting evidence;

interpret legislation;

determine the remedy.

AI should therefore remain an instrument of judicial administration or assistance, unless a specific legal framework provides otherwise.

42. Transparency Does Not Mean Automation Must Be Abandoned

The existence of explainability requirements does not mean that AI cannot be used.

AI can significantly assist courts by:

reducing repetitive work;

organising large records;

identifying relevant material;

improving translation;

calculating financial claims;

supporting digital case management.

The legal requirement is better expressed as:

Use technology, but preserve accountable human reasoning.

43. Future UAE Legal Development

A mature UAE framework for automated judgments could potentially establish:

mandatory disclosure of material AI use;

human judicial sign-off;

audit trails;

model validation;

bias testing;

data-quality controls;

explanation requirements;

correction mechanisms;

appeal access;

judicial AI registers;

independent technological audits;

protection of confidential algorithms;

standards for AI-generated evidence;

clear responsibility for system errors.

Such reforms could create a balance between:

innovation + efficiency + fairness + accountability.

44. Key Legal Principles

Principle 1

There is presently no general UAE statutory provision expressly creating a standalone "right to explanation" for AI-generated judgments.

Principle 2

The requirement for reasoned judgments provides a strong foundation for explainability.

Principle 3

Article 129 of the Civil Procedure Code requires judgments to state their underlying grounds.

Principle 4

Article 130 requires the judgment to contain facts, claims, defences, grounds and operative provisions.

Principle 5

AI output cannot automatically substitute for judicial reasoning.

Principle 6

Material evidence and defences must receive appropriate judicial consideration.

Principle 7

Technical experts cannot replace the court's legal determination.

Principle 8

AI-generated legal material must be verified.

Principle 9

The DIFC Courts expressly emphasise transparency, accuracy, reliability and human decision-making in their AI guidance.

Principle 10

Meaningful explanation is essential for effective appellate review.

45. Suggested Legal Model

The UAE's emerging model can be represented as:

Digital Data

↓

AI/Algorithmic Processing

↓

Technical/Analytical Output

↓

Human Judicial Verification

↓

Party Challenge

↓

Evaluation of Evidence

↓

Application of Law

↓

Reasoned Judgment

↓

Appeal/Cassation

This model preserves technological efficiency without converting the algorithm into an unaccountable legal authority.

46. Conclusion

The right to explanation in automated judgments is an emerging principle in UAE civil law rather than an already codified standalone right.

Its legal foundation comes from the established requirement that judgments be properly reasoned, that material evidence and defences be considered, that technical expertise not replace judicial determination, and that parties have meaningful avenues for review.

The strongest contemporary authority is Oheo Bank v Parker [2025] DIFC CA 006, where the DIFC Court of Appeal stressed that adequate reasons are essential to effective appellate review and that a series of unexplained conclusions is insufficient.

The developing UAE approach can therefore be expressed as:

AI may assist the court, but the judgment must remain a humanly attributable, legally reasoned and reviewable judicial decision.

A legally meaningful automated judgment should allow the affected party to understand:

what was decided → what evidence mattered → what law was applied → what arguments were considered → why the conclusion followed → how the decision can be challenged.

That is the practical essence of a right to explanation in automated civil adjudication.

47. Quick Revision Notes

Definition

Right to explanation = ability to understand the legally significant basis of an automated or AI-assisted decision sufficiently to challenge and review it.

Main UAE statutory foundation

Federal Decree-Law No. 42 of 2022 – Civil Procedure Code.

Article 129 – underlying grounds of judgments.

Article 130 – facts, claims, defences, grounds and operative part.

Federal Decree-Law No. 35 of 2022 – Law of Evidence.

Important technology framework

DIFC Digital Economy Court.

DIFC Practical Guidance Note No. 2 of 2023 on LLMs and generative AI.

Core principle

AI assistance is not a substitute for judicial reasoning.

Key cases

Oheo Bank v Parker [2025] DIFC CA 006 – adequate reasons and effective appellate review.

Civil Cassation No. 647/2021 – material evidence and substantial defences.

Commercial Cassation No. 215/2020 – reasoned assessment of expert evidence.

Commercial Cassation No. 767/2021 – technical expertise does not determine legal questions.

Commercial Cassation No. 872/2023 – clear and sufficient judicial reasons.

Commercial Cassation Nos. 1012 & 1023/2022 – technical analysis cannot replace legal determination.

Arabyads Holding Ltd v Gulrez Alam Marghoob Alam [2025] ADGMCFI 0032 – human verification of AI-generated legal material.

Johnson Arabia LLC v BIC Contracting LLC [2020] DIFC CFI 075 – identification of material issues and proper judicial analysis.

Exam formula

AI output → human review → evidence → law → reasoning → judgment → appeal

rather than:

AI output → automatic judgment.