1. Introduction

A Human–AI Hybrid Justice Model is a judicial system in which artificial intelligence assists human judges and court officials, while the final legal authority and responsibility remain with human decision-makers.

The model can be expressed simply as:

AI assists → Human verifies → Judge evaluates → Human judge decides → Court gives reasoned judgment

This is different from a fully automated justice model, where an algorithm would independently determine rights and liabilities.

For UAE civil law, the hybrid model is particularly relevant because the UAE has rapidly developed:

• electronic litigation;
• electronic evidence;
• digital signatures;
• automated transactions;
• online dispute-resolution mechanisms;
• AI-assisted legal research;
• digital courts;
• blockchain and digital-asset systems; and
• AI-assisted drafting and case-management technologies.

However, there is still a major distinction between technological assistance and the exercise of judicial authority.

A careful legal point is necessary: reported UAE mainland cases directly deciding whether AI itself may adjudicate a civil dispute remain very limited. Consequently, the case law below combines the emerging UAE/DIFC/ADGM AI authorities with UAE-region decisions concerning digital evidence, technological assets, jurisdiction and human judicial assessment.

2. Meaning of a Human–AI Hybrid Justice Model

A hybrid justice model has two components.

Human component

Humans retain responsibility for:

• determining disputed facts;
• interpreting legislation;
• evaluating credibility;
• applying legal principles;
• exercising judicial discretion;
• determining liability;
• determining remedies; and
• issuing the final judgment.

AI component

AI may assist with:

• searching case law;
• organising documents;
• identifying relevant evidence;
• summarising lengthy records;
• detecting inconsistencies;
• translating documents;
• transcription;
• case classification;
• procedural scheduling;
• preparing preliminary drafts;
• analysing large datasets; and
• identifying potentially relevant legal provisions.

Therefore:

AI is a judicial assistant, not an autonomous judicial authority.

3. UAE Legal Foundation

A. Civil Transactions Law

The UAE's new Civil Transactions Law, Federal Decree-Law No. 25 of 2025, effective from 1 June 2026, modernises the general framework of UAE civil law.

For AI-assisted justice, its importance lies in the fact that technologically generated information must ultimately be evaluated through ordinary civil-law concepts such as:

• contractual obligation;
• good faith;
• causation;
• fault;
• damage;
• unjust enrichment;
• abuse of rights;
• liability;
• interpretation of agreements; and
• judicial discretion.

Technology does not eliminate these legal concepts.

4. Electronic Transactions and Trust Services Law

Federal Decree-Law No. 46 of 2021 recognises electronic transactions and automated electronic processes.

The legislation deals with:

• electronic documents;
• electronic signatures;
• electronic records;
• electronic communications;
• electronic identification;
• trust services; and
• automated electronic transactions.

This is important because the UAE legal system already accepts that certain legal consequences can arise through automated systems.

But there is an important distinction:

Automated transaction

A computer system performs a transaction according to programmed rules.

Automated judgment

A computer determines the legal rights and liabilities of disputing parties.

The first is expressly accommodated by modern electronic-commerce legislation. The second raises substantially greater questions of judicial authority, procedural fairness and accountability.

5. UAE Evidence Law

Federal Decree-Law No. 35 of 2022 on Evidence in Civil and Commercial Transactions expressly recognises electronic evidence.

Electronic evidence includes information generated, stored, extracted, copied, transmitted or received through information technology and capable of being retrieved in understandable form.

This permits courts to deal with:

• emails;
• electronic records;
• databases;
• digital communications;
• system logs;
• electronic signatures;
• electronic documents;
• recordings;
• digital photographs; and
• other electronically generated information.

The important principle is:

Electronic evidence may be processed technologically, but its legal weight remains a matter for judicial assessment.

6. The Basic Hybrid Justice Architecture

A UAE human–AI justice system could operate as follows:

1. Digital evidence

↓

2. AI processing

↓

3. AI recommendation

↓

4. Human verification

↓

5. Parties' opportunity to challenge

↓

6. Judicial evaluation

↓

7. Human judicial decision

↓

8. Reasoned judgment

This model maintains technological efficiency without transferring judicial responsibility to an algorithm.

7. Case Law 1 — Arabyads Holding Limited v Gulrez Alam Marghoob Alam

[2025] ADGMCFI 0032 — ADGM

This is currently one of the most significant UAE-related authorities concerning AI-assisted legal work.

The case concerned legal submissions prepared with substantial assistance from AI. The material contained false or incorrectly cited legal authorities. The ADGM Court considered whether the lawyers responsible for the AI-assisted work should bear wasted-cost consequences.

The Court emphasised the professional obligation to verify AI-generated legal research before relying upon it.

The Court ultimately imposed substantial costs consequences on the relevant legal representatives.

Principle

AI assistance does not transfer professional responsibility from the human user to the machine.

Relevance to hybrid justice

This principle has an even stronger application to judicial AI.

If a lawyer must verify AI-generated legal material, a judicial institution should likewise require verification before an AI-generated:

• case summary;
• legal proposition;
• evidentiary analysis;
• draft judgment; or
• recommendation

is relied upon.

Formula

AI output → human verification → legal responsibility

Jurisdictional caution

ADGM is a separate common-law jurisdiction. This is not a binding mainland UAE Court of Cassation decision, but it is highly relevant to the UAE's developing approach to AI-assisted legal processes.

8. Case Law 2 — Klesta Eshja & Hair Creators Salon LLC v Salah Masri & Others

DIFC CFI 066/2024

This is another highly relevant recent DIFC authority.

The defendants' amended defence submissions were found to have been prepared substantially with AI assistance and to contain false references and misleading material. The DIFC Court ordered the problematic pleadings to be struck out and subsequently dealt with consequential costs.

The case demonstrates that AI-generated litigation material is not immune from ordinary procedural requirements.

Principle

A litigant cannot avoid procedural responsibility merely because a document was generated or assisted by AI.

Hybrid-justice significance

The case establishes a useful conceptual rule:

The legal responsibility attaches to the human actor using the technology, not to the algorithm.

If this principle applies to litigants, a similar governance principle should apply when courts use AI.

The judicial officer must remain responsible for verifying AI-assisted outputs.

9. Case Law 3 — VTB Bank PJSC v Kuanyshev & Others

DIFC CFI 121/2025

This is particularly relevant to AI-generated legal reasoning.

The DIFC Court encountered pleadings containing what appeared to be AI-generated arguments and numerous fabricated or incorrectly described authorities. The Court identified apparent hallucinations in the legal material and expressly referred to the DIFC's existing guidance concerning the use of generative AI.

Principle

Plausible-looking legal reasoning is not necessarily legally reliable.

An AI system can generate:

• realistic case names;
• plausible legal propositions;
• incorrect statutory provisions;
• nonexistent authorities; and
• apparently logical but legally incorrect arguments.

Hybrid-justice significance

A human judge must therefore be able to ask:

1. Does the authority exist?
2. Does the case actually say what the AI claims?
3. Is the legislation current?
4. Is the authority applicable to the jurisdiction?
5. Has the AI confused one legal system with another?
6. Does the evidence actually support the conclusion?

This is a central feature of a responsible human–AI judicial model.

10. Case Law 4 — Gate Mena DMCC v Tabarak Investment Capital Ltd & Christian Thurner

[2023] DIFC CA 002

The Gate Mena litigation is particularly important for technologically complex civil disputes.

The dispute involved cryptocurrency, Bitcoin, digital wallets, private keys and blockchain technology.

The DIFC Court of Appeal held that Bitcoin was capable of constituting property under the applicable DIFC legal framework and dealt with complex technological and legal questions surrounding digital assets.

Principle

Technological facts require legal interpretation.

A blockchain may show that a transaction occurred, but the court still has to determine:

• ownership;
• control;
• possession;
• contractual obligations;
• causation;
• liability; and
• appropriate remedies.

Hybrid-justice significance

An AI system might successfully identify thousands of blockchain transactions.

But it cannot simply convert:

“Transaction occurred”

into:

“Defendant is legally liable.”

The human judge must make that legal connection.

11. Case Law 5 — DNB Bank ASA v Gulf Eyadah Corporation & Gulf Navigation Holding PJSC

[2015] DIFC CA 007

This leading DIFC decision concerned recognition and enforcement of an English judgment.

The Court examined:

• jurisdiction;
• foreign judgments;
• applicable legal rules;
• recognition;
• enforcement; and
• the legal effect of the foreign judgment.

Principle

Jurisdiction and legal effect are judicial questions.

An automated system could identify that:

• a foreign judgment exists;
• a judgment debt exists;
• the parties are connected;
• assets may be present.

But it cannot automatically determine the legal consequence without applying the applicable legal rules.

Hybrid-justice application

An AI system might produce:

“Foreign judgment satisfies enforcement criteria.”

The judge must still independently verify:

• jurisdiction;
• applicable law;
• procedural requirements;
• public-policy considerations;
• finality;
• enforceability; and
• relief.

Thus:

AI can classify; the judge must adjudicate.

12. Case Law 6 — DNB Bank ASA v Gulf Eyadah Corporation

[2014] DIFC CFI 043

The first-instance proceedings in DNB Bank similarly required detailed judicial consideration of the DIFC Court's jurisdiction to recognise and enforce the foreign judgment.

The Court considered the jurisdictional gateways and the interaction of the applicable legal framework.

Principle

Legal disputes involving multiple jurisdictions cannot safely be reduced to simple automated classifications.

Hybrid-justice significance

A judicial AI system could assist by identifying:

• foreign judgments;
• jurisdictional clauses;
• relevant statutory provisions;
• prior cases.

But the judge must determine:

Which legal rule actually governs this particular dispute?

This is especially important in UAE's multi-layered legal environment involving:

• federal courts;
• Dubai Courts;
• DIFC Courts;
• ADGM Courts;
• arbitration;
• foreign judgments; and
• international conventions.

13. Case Law 7 — Gate Mena Digital-Asset Proceedings

TCD 001/2020

The first-instance Gate Mena proceedings concerned a technologically complex dispute involving cryptocurrency and digital assets.

The Court had to consider technical evidence concerning digital wallets, private keys, control and the operation of cryptocurrency systems.

Principle

Technical expertise can assist the court without replacing the court.

Hybrid model

The proper relationship is:

Technical evidence

AI/data analysis

Expert explanation

↓

Human judicial determination

The judge remains responsible for determining the legal significance of the technical facts.

14. Case Law 8 — Stelian Gheorghe v BSA Ahmad Bin Hezeem & Associates LLP & Jimmy Haoula

DIFC CFI 045/2025

This emerging DIFC litigation also illustrates the developing judicial concern with AI-generated or AI-assisted litigation material.

The importance of this line of cases is not that AI is automatically prohibited.

Rather, the courts are increasingly concerned with whether material generated through AI:

• is accurate;
• contains genuine authorities;
• satisfies procedural requirements;
• is properly verified; and
• can safely be relied upon.

Principle

Technology must remain subordinate to procedural and professional obligations.

This is a central concept for a hybrid justice model.

15. DIFC Practical Guidance on Generative AI

The DIFC Courts have gone further than simply dealing with individual cases.

Their Practical Guidance Note No. 2 of 2023 addresses the use of large language models and generative AI in court proceedings.

The guidance identifies risks including:

• incorrect information;
• misleading evidence;
• confidentiality breaches;
• intellectual-property concerns;
• data-protection issues;
• bias;
• excessive reliance on AI.

It emphasises:

• transparency;
• accuracy;
• reliability;
• verification;
• early disclosure;
• confidentiality; and
• avoiding over-reliance on generative AI.

Most importantly for hybrid justice, it says AI should assist rather than replace the integral human decision-making required in preparing material for court.

16. What Is Actually "Hybrid" About the Model?

A hybrid justice system does not mean that every task is divided 50/50 between humans and AI.

Instead, different functions may be allocated according to their suitability.

17. Human-in-the-Loop Model

The most important model is the human-in-the-loop system.

Level 1 — AI processing

AI processes large amounts of information.

Level 2 — AI recommendation

AI identifies potentially relevant conclusions.

Level 3 — Human review

A judicial officer checks the output.

Level 4 — Judicial reasoning

The judge independently applies the law.

Level 5 — Human decision

The judge issues the decision.

This prevents automation from becoming a substitute for judicial authority.

18. Human-on-the-Loop Model

A weaker form of oversight is the human-on-the-loop model.

Here:

• AI performs substantial processing;
• humans monitor the system;
• humans can intervene where necessary.

This may be suitable for:

• scheduling;
• administrative classification;
• document management;
• duplicate detection;
• case-routing systems.

However, it is less appropriate for the final determination of contested civil rights.

19. Human-in-Command Model

For the most significant judicial decisions, a stronger model is preferable:

Human-in-command

Under this model:

• AI cannot issue the final judgment;
• AI cannot determine liability independently;
• AI cannot impose civil liability autonomously;
• AI cannot independently determine damages;
• AI cannot prevent judicial reconsideration;
• the judge can reject the AI recommendation.

This preserves judicial accountability.

20. AI-Assisted Judicial Drafting

AI may potentially assist a judge by producing a preliminary draft containing:

• facts;
• procedural history;
• disputed issues;
• relevant provisions;
• previous authorities;
• possible reasoning.

But the judge should verify:

Facts

Are they actually supported by the record?

Law

Are the provisions current?

Authorities

Do the cases exist?

Reasoning

Does the conclusion logically follow?

Remedy

Is the remedy legally available?

Parties

Have all material arguments been considered?

The final judgment should therefore remain a human-authored judicial determination, even if technology assisted its preparation.

21. AI and Judicial Discretion

This is particularly important in civil law.

Judges frequently have to determine:

• good faith;
• abuse of rights;
• reasonable conduct;
• causation;
• foreseeability;
• mitigation;
• hardship;
• damages;
• proportionality;
• contractual intention.

These concepts involve contextual judgment.

For example:

An AI system may calculate that a contractual breach caused 80% of a loss.

But the judge must still decide whether the legal requirements of causation are satisfied.

Therefore:

Statistical correlation ≠ legal causation.

22. AI and Evidence

AI can help organise evidence but should not automatically determine evidentiary weight.

Consider an email.

AI may identify:

• date;
• sender;
• recipient;
• keywords;
• sentiment;
• references to a contract.

But the judge still has to determine:

• authenticity;
• attribution;
• context;
• relevance;
• reliability;
• probative value.

This follows the broader logic of the UAE Evidence Law.

23. AI and Expert Evidence

AI may also assist experts.

For example:

Construction dispute

AI analyses:

• thousands of project records;
• BIM data;
• construction schedules;
• invoices;
• correspondence.

The expert then evaluates the technical results.

The judge then evaluates:

• the expert's methodology;
• competing expert opinions;
• contractual provisions;
• causation;
• damages.

Thus:

AI → Expert → Judge

rather than:

AI → Judgment

24. AI and Smart Contracts

The UAE's electronic-transactions framework is compatible with automated transactions.

A smart contract might automatically:

• release payment;
• transfer a digital asset;
• calculate a charge;
• trigger an obligation.

But if a dispute occurs, the court may have to determine:

• whether the smart contract was valid;
• whether consent existed;
• whether fraud occurred;
• whether the automated event was triggered correctly;
• whether a force-majeure event existed;
• whether damages are recoverable.

AI may help analyse the code and transaction history.

The judge determines the legal consequences.

25. AI and Digital Assets

The Gate Mena litigation demonstrates why this is important.

Digital assets may involve:

• blockchain;
• wallets;
• private keys;
• smart contracts;
• automated transactions.

AI can process these systems faster than humans.

However, the legal questions remain human questions:

Who owns the asset?

Who controlled it?

Was there a breach of duty?

Was there fraud?

What remedy is available?

26. AI and Procedural Fairness

A hybrid justice system must protect procedural fairness.

Parties should have appropriate opportunities to know and challenge material that significantly affects the decision.

This becomes particularly important if an AI system:

• identifies evidence;
• filters evidence;
• classifies claims;
• recommends legal authorities;
• identifies likely outcomes.

A party should not effectively lose a civil case because of an opaque algorithm that cannot be meaningfully challenged.

27. AI and Explainability

Explainability means that the court should be able to understand the basis of an AI recommendation sufficiently to assess it.

For example:

Weak system

"The claimant has a 73% probability of success."

Better system

"The system identified these contractual provisions, these factual inputs and these authorities as relevant."

Even then, the judge must decide independently.

The objective is not necessarily to reveal every line of source code.

The objective is to make the legally significant reasoning and evidentiary basis sufficiently understandable and reviewable.

28. AI and Algorithmic Bias

AI can reproduce bias contained in training data.

For example, if historical judicial data contains patterns that disproportionately affect a particular category of litigants, an AI system trained on that data could reproduce those patterns.

Human oversight should therefore include:

• bias testing;
• data auditing;
• independent review;
• monitoring of unusual outcomes;
• periodic evaluation.

The judge should not assume:

"The computer is neutral because it is mathematical."

Mathematical processing can reproduce human assumptions embedded in the data.

29. Automation Bias

There is also a risk on the human side.

A judge may unconsciously trust an AI recommendation because it appears:

• objective;
• mathematical;
• sophisticated;
• data-driven.

This is called automation bias.

Human oversight therefore means more than simply putting a person somewhere in the process.

The human must have:

• genuine authority;
• sufficient information;
• ability to question the system;
• ability to reject its recommendation.

30. AI Audit Trail

A proper hybrid justice system should maintain an audit trail recording:

• which AI system was used;
• when it was used;
• what data was supplied;
• what output was produced;
• what version of the model was used;
• who reviewed the output;
• what changes were made;
• whether the judge rejected or accepted the recommendation.

This creates accountability.

31. Human Override

A fundamental requirement should be:

The human decision-maker must be able to override the AI recommendation.

For example:

AI recommends:

Reject claim.

Judge determines:

The algorithm failed to consider an exceptional contractual circumstance.

The judge must be able to reject the recommendation without the system preventing or penalising that decision.

32. No "Black Box Judge"

A fundamental legal danger is the creation of a black-box judge.

A black-box system would:

• process evidence;
• apply hidden criteria;
• produce an outcome;
• provide little explanation;
• prevent meaningful challenge.

Such a system would create serious concerns concerning:

• due process;
• judicial accountability;
• transparency;
• equality of arms;
• appeal;
• reasoned judgments.

The hybrid model should therefore reject:

Black-box automation of final civil adjudication.

33. Relationship Between AI and Judicial Independence

Judicial independence requires that judges be able to make decisions according to law.

If an AI recommendation becomes practically impossible to reject because:

• institutional policy requires compliance;
• the system controls case allocation;
• performance metrics depend on AI recommendations;
• judges cannot understand the model;

then formal human involvement may exist without genuine human control.

Therefore:

Human oversight must be substantive, not merely symbolic.

34. Multi-Level UAE Justice and AI

The UAE's judicial environment is legally diverse.

It includes:

• Federal Courts;
• Dubai Courts;
• Abu Dhabi Courts;
• DIFC Courts;
• ADGM Courts;
• arbitration tribunals.

An AI system therefore must not confuse legal authorities from different jurisdictions.

For example:

DIFC case ≠ automatically binding mainland UAE precedent.

ADGM case ≠ automatically binding Dubai Court precedent.

This is particularly important because AI systems may combine cases from multiple databases without understanding jurisdictional hierarchy.

35. The VTB Problem: AI Hallucination and Jurisdictional Confusion

The VTB litigation is particularly instructive because the Court itself identified apparent AI-style hallucinations and incorrect authorities.

This demonstrates a serious problem:

An AI system can produce a perfectly plausible statement such as:

"The DIFC Court previously held X in Case Y."

But:

• Case Y may not exist;
• Case Y may say something else;
• the case may be from another jurisdiction;
• the case may have been overruled;
• the legislation may have changed.

Human legal verification is therefore essential.

36. Six Core Case Laws — Revision Table

Important: These are not six precedents establishing a single binding UAE mainland rule on AI judges. The first three are particularly relevant to AI-assisted litigation; the Gate Mena and DNB authorities illustrate broader principles of technological evidence, jurisdiction and human legal judgment. The distinction between mainland UAE, DIFC and ADGM jurisdiction is essential.

37. Difference Between AI-Assisted and AI-Decided Justice

38. Advantages of Human–AI Hybrid Justice

1. Efficiency

AI can process enormous volumes of material.

2. Faster research

Relevant legislation and case law can be identified quickly.

3. Document management

Thousands of documents can be classified and searched.

4. Consistency

AI can identify potentially inconsistent treatment or reasoning.

5. Access to justice

Digital tools can make legal information more accessible.

6. Reduced administrative burden

Judges can spend more time on genuinely disputed questions.

39. Risks

1. Hallucinations

AI can invent authorities.

2. Bias

Historical data can influence outputs.

3. Automation bias

Humans may overtrust AI.

4. Privacy

Judicial records may contain sensitive information.

5. Cybersecurity

Manipulation of judicial systems could have serious consequences.

6. Explainability

Some AI models are difficult to explain.

7. Accountability

Responsibility may become unclear if governance is poorly designed.

40. Recommended UAE Hybrid Justice Framework

A strong framework should contain:

A. Human authority

Only legally authorised judges issue final judgments.

B. AI verification

AI-generated legal material must be verified.

C. Transparency

Material use of AI should be appropriately documented.

D. Explainability

Significant recommendations should be sufficiently understandable.

E. Auditability

AI activity should be recorded.

F. Human override

Judges must be able to reject AI outputs.

G. Data protection

Judicial data must be protected.

H. Bias monitoring

Models should be periodically tested.

I. Procedural challenge

Parties should have appropriate opportunities to challenge material AI-assisted evidence.

J. Jurisdictional controls

AI databases must distinguish:

• Federal UAE law;
• Dubai law;
• DIFC law;
• ADGM law;
• foreign law;
• arbitration decisions.

41. Three-Level Human Control Model

A useful examination framework is:

Level 1 — Operational oversight

Humans supervise:

• data;
• software;
• security;
• system operation.

Level 2 — Evidentiary oversight

Humans verify:

• documents;
• evidence;
• AI classifications;
• factual conclusions.

Level 3 — Judicial oversight

The judge independently determines:

• law;
• facts;
• liability;
• remedies;
• final judgment.

The third level should never be completely surrendered to AI.

42. Central Legal Principle

The emerging UAE approach can be summarised as:

Automation may support adjudication, but automation should not replace judicial responsibility.

The distinction is particularly important because the UAE is technologically ambitious while maintaining a structured judicial hierarchy.

The appropriate objective is therefore not:

Human versus AI

but:

Human + AI, with clearly defined legal responsibility.

43. Short Exam Answer

Human–AI hybrid justice in UAE civil law refers to a system where artificial intelligence assists courts with legal research, evidence organisation, document analysis, translation, case management and preliminary drafting, while human judges retain responsibility for legal interpretation and the final decision.

The UAE's Electronic Transactions and Trust Services Law recognises automated electronic transactions, while the Evidence Law recognises electronic evidence. However, these technological frameworks do not mean that an AI system independently possesses judicial authority.

Recent UAE-region cases demonstrate the importance of human verification. Arabyads Holding v Alam [2025] ADGMCFI 0032 concerned AI-assisted legal research containing false or incorrectly cited authorities. Klesta Eshja v Masri, CFI 066/2024, and VTB Bank v Kuanyshev, CFI 121/2025, similarly demonstrate the procedural dangers of unverified AI-generated litigation material. Gate Mena v Tabarak demonstrates the need for judicial interpretation of technologically complex digital assets, while DNB Bank v Gulf Eyadah demonstrates that jurisdiction and legal effect remain judicial questions.

Thus, the appropriate model is:

AI processing → AI recommendation → human verification → judicial reasoning → human judgment.

Conclusion

The Human–AI Hybrid Justice Model represents a middle path between traditional manual adjudication and fully automated justice.

For UAE civil law, the most defensible model is one in which:

AI provides speed and analytical assistance, while humans retain legal authority, judgment, accountability and control.

The emerging Arabyads, Klesta Eshja and VTB Bank authorities demonstrate the dangers of treating AI-generated legal material as automatically reliable. Gate Mena demonstrates the importance of human interpretation of technologically complex evidence, while DNB Bank demonstrates why jurisdiction and legal consequences require judicial determination.

The fundamental formula for UAE automated justice can therefore be remembered as:

AI assists — humans verify — judges reason — courts decide.

Exam keywords: human-in-the-loop, human-in-command, AI-assisted adjudication, judicial accountability, explainability, verification, electronic evidence, automation bias, algorithmic bias, procedural fairness, human override, audit trail, digital justice.