Civil Law and UAE: Machine Learning Governance of Legal Systems

1. Introduction

Machine learning governance of legal systems concerns the legal rules, institutional safeguards and judicial principles governing the use of machine-learning and artificial-intelligence systems in:

• courts;
• legal research;
• evidence management;
• case classification;
• fraud detection;
• regulatory supervision;
• judicial administration;
• automated dispute-resolution systems;
• digital evidence analysis;
• legal-document generation;
• prediction and risk assessment.

The UAE is increasingly developing a legal environment in which technology can assist legal institutions. The DIFC Digital Economy Court, for example, expressly covers disputes involving AI, digital assets, blockchain, complex databases, cloud data and digital platforms, and its rules permit AI-driven smart forms and decision-tree software.

At the same time, the DIFC Courts have expressly warned that AI-generated material can create risks of incorrect information, confidentiality breaches, intellectual-property problems and data-protection violations, and have emphasised verification, transparency and human decision-making.

Therefore, the central principle is:

Machine learning may assist the legal system, but legal authority, procedural fairness and judicial responsibility cannot automatically be transferred to an algorithm.

2. Meaning of Machine Learning Governance

Machine learning governance means establishing rules for:

1. what AI systems may be used for;
2. what data they may use;
3. how they are tested;
4. how their outputs are verified;
5. who is responsible for errors;
6. how bias is identified;
7. how decisions are explained;
8. how confidential information is protected;
9. how affected persons can challenge automated results;
10. how human judicial control is maintained.

It is therefore broader than merely regulating software.

A useful formula is:

ML Governance = Legality + Data Governance + Transparency + Accuracy + Accountability + Human Oversight + Security + Reviewability

3. Why Machine Learning Creates a Civil-Law Problem

Traditional civil adjudication generally involves:

Law → Facts → Evidence → Interpretation → Judgment

Machine-learning systems introduce another layer:

Data → Model → Prediction/Classification → Human or Automated Decision

This raises difficult questions.

Example

Suppose an algorithm predicts that a particular contractual claim has only a 20% probability of succeeding.

Can the judge simply reject the claim?

No—not merely because the algorithm produced that percentage.

The court must still consider:

• applicable law;
• admissible evidence;
• factual findings;
• arguments of the parties;
• procedural rights;
• reasoning;
• judicial authority.

The algorithm provides analytical assistance, not an independent source of legal authority.

4. UAE Legal Environment

The UAE does not currently have one single comprehensive federal statute titled a "Machine Learning Governance of Courts Act."

Instead, governance is distributed across several legal and institutional frameworks, including:

• civil and procedural law;
• evidence law;
• personal-data protection;
• electronic transactions and trust services;
• sector-specific regulation;
• cybersecurity requirements;
• judicial rules;
• DIFC legislation and court rules;
• ADGM regulatory frameworks;
• professional obligations.

The DIFC framework is particularly developed in relation to digital disputes.

Part 58 of the DIFC Courts Rules permits Digital Economy Court claims concerning AI and AI-controlled devices and expressly permits the Court to conduct proceedings digitally. It also allows electronic AI-driven forms based on decision-tree software.

5. Human Judicial Authority

The most important governance principle is:

AI should support judicial reasoning rather than replace judicial authority.

The DIFC Courts' Practical Guidance Note No. 2 of 2023 expressly states that LLMs and generative AI should assist parties and should not replace the integral human decision-making required in preparing evidence and submissions.

The same principle is even more important where an AI system is used directly within a legal institution.

A machine-learning model may:

• identify similar cases;
• classify documents;
• detect anomalies;
• organise evidence;
• estimate financial consequences.

But the judge must retain responsibility for:

• interpreting law;
• determining disputed facts;
• evaluating evidence;
• applying legal standards;
• deciding credibility;
• determining liability;
• determining remedies.

6. Principle of Legality

Machine learning cannot independently create a legal rule simply because a model has detected a pattern.

For example:

The model observes that courts usually award AED 100,000 in a particular category of dispute.

That does not automatically establish:

AED 100,000 is legally mandatory.

A civil-law system places legislation and recognised legal sources at the centre of legal authority.

Thus:

Statistical regularity ≠ legal rule.

This is especially important for the UAE's codified legal system.

7. Data Governance

Machine learning is only as reliable as its data environment.

Legal AI may process:

• judgments;
• contracts;
• witness statements;
• financial records;
• personal information;
• medical evidence;
• emails;
• digital communications;
• transaction records.

Governance therefore requires:

Data accuracy

Is the information correct?

Data relevance

Is it legally relevant?

Data completeness

Are important documents missing?

Data provenance

Where did the data originate?

Data integrity

Has it been altered?

Data security

Who can access it?

Data retention

How long should it remain available?

8. Personal Data and Machine Learning

Machine-learning legal systems can process large amounts of personal data.

This creates risks involving:

• privacy;
• confidentiality;
• profiling;
• unauthorised disclosure;
• cross-border transfers;
• excessive data collection.

The DIFC Courts' AI guidance expressly requires practitioners using generative AI to consider the DIFC Data Protection Law 2020 and client confidentiality.

The same principle is relevant to machine-learning systems used by legal institutions:

Judicial efficiency cannot eliminate data-protection obligations.

9. Transparency

An affected party should, where legally appropriate, be able to understand the role played by AI in a decision.

Transparency may include:

• whether AI was used;
• what function it performed;
• what data category it relied upon;
• whether a human reviewed the result;
• what limitations were known;
• whether the output was independently verified.

The DIFC AI guidance expressly identifies transparency as a core principle and calls for disclosure of AI-generated content and relevant limitations or biases.

10. Explainability

Machine-learning models can be difficult to explain, particularly complex models.

This creates the "black box" problem:

Input → Unknown internal process → Output

In legal adjudication, that may be insufficient.

A judicial decision ordinarily needs a reasoned legal explanation.

Therefore:

Explainability is not merely a technical preference; it can be connected to procedural fairness and meaningful judicial review.

11. Accuracy and Verification

The DIFC Courts specifically require AI-generated content to be verified for accuracy and reliability before it is relied upon. The guidance warns against reliance on unverified AI outputs and notes that courts can reject AI-generated content under the relevant rules.

This produces a basic rule:

AI output is evidence or assistance—not automatically truth.

For machine-learning systems, verification may involve:

• independent source checking;
• expert review;
• testing;
• validation datasets;
• error analysis;
• comparison against primary legal sources.

12. Bias and Equality

Machine learning can reproduce biases contained in training data.

For example, if historical decisions contain an institutional bias, a model trained on those decisions may reproduce the pattern.

The governance problem is:

Should historical frequency determine future legal treatment?

The answer cannot simply be "yes."

A court must apply the law to the particular dispute.

Machine learning therefore requires:

• bias testing;
• representative data;
• fairness monitoring;
• human review;
• challenge mechanisms.

13. Procedural Fairness

A party should not ordinarily be disadvantaged merely because an undisclosed algorithm classified the case in a particular way.

Procedural fairness may require:

• notice;
• opportunity to respond;
• access to relevant evidence;
• impartial decision-maker;
• reasoned determination;
• ability to challenge an adverse result.

This becomes particularly important if machine learning is used for:

• case triage;
• risk classification;
• fraud detection;
• automated dispute resolution;
• evidence prioritisation.

14. Case Law 1 — Gate Mena DMCC v Tabarak Investment Capital Ltd [2024] DIFC DEC 002

This is one of the most important recent UAE cases for digital judicial governance.

The case was heard by the DIFC Digital Economy Court and concerned cryptocurrency transactions, digital assets and questions concerning control and transfer of Bitcoin.

The Court's June 2026 judgment arose after an earlier Court of Appeal decision ordered a retrial of a particular issue and remitted the matter to the Digital Economy Court.

Governance significance

The case demonstrates that legal institutions must adapt established legal concepts to technologically complex environments without abandoning ordinary judicial methodology.

The court still had to determine:

• facts;
• evidence;
• legal rights;
• ownership/control;
• contractual obligations.

Principle

Technological complexity does not eliminate ordinary requirements of legal reasoning and proof.

For machine-learning governance, this means an AI system should assist the court in analysing technological evidence but cannot independently determine its legal significance.

15. Case Law 2 — Graciela Limited v Giacobbe [2014] DIFC CFI 027

This is an important technology-evidence case.

The defendant was alleged to have deliberately sabotaged the claimant's IT system.

The Court considered:

• server information;
• IP addresses;
• user accounts;
• system logs;
• forensic analysis;
• expert evidence.

The Court accepted strong circumstantial and technical evidence establishing responsibility and awarded compensatory damages.

Most importantly, the judge expressly stated that the court—not the expert—must decide the ultimate question.

Machine-learning governance principle

This provides a useful analogy:

Technical expertise may inform the court, but technical systems do not determine the legal conclusion.

That principle is highly relevant to machine-learning evidence.

16. Case Law 3 — Aegis Resources DMCC v Union Bank of India [2020] DIFC CFI 004

This case concerned cyber fraud involving fraudulent payment instructions sent through a compromised email system.

The Court had to determine:

• how the fraud occurred;
• who bore the risk;
• what security obligations existed;
• whether the bank or customer should bear the loss.

The Court described the matter as an emerging cyber-fraud dispute and concluded, on the particular facts, that the loss fell on the bank, with some consequential loss recoverable by the customer.

Machine-learning governance significance

The case demonstrates that responsibility for technology-related harm is fact-specific.

An automated governance system therefore cannot simply apply:

“Cyber incident = customer responsibility.”

It must examine:

• system architecture;
• contractual obligations;
• security practices;
• causation;
• conduct of the parties.

17. Case Law 4 — Lals Holdings Ltd v Emirates Insurance Company [2024] DIFC CA 002

The case involved business-interruption insurance claims arising from COVID-19 and alternative claims against an insurance broker.

The DIFC Court of Appeal examined the interpretation of insurance policies and the relationship between the insured, insurer and broker.

Machine-learning governance significance

This demonstrates why an algorithm cannot interpret contractual language merely by statistical frequency.

The court must consider:

• contractual text;
• context;
• commercial purpose;
• applicable legal principles;
• surrounding circumstances.

Thus:

Machine-learning pattern recognition cannot automatically substitute for legal interpretation.

18. Case Law 5 — Al Ramz Capital LLC v DFSA [2025] DIFC CFI 087

This case concerned regulatory enforcement involving alleged market-abuse reporting failures.

The dispute involved questions concerning:

• regulatory obligations;
• market surveillance;
• suspected wash trading;
• reporting obligations;
• privacy/publication of regulatory decisions.

The DIFC Court refused permission to appeal because the proposed grounds did not establish an appealable error of law.

Machine-learning governance significance

Financial institutions increasingly use machine-learning systems to detect suspicious trading.

The case illustrates an important principle:

Technology may assist regulatory surveillance, but the legal obligation remains with the regulated institution.

A financial institution cannot simply say:

“Our algorithm did not detect it.”

If the law imposes a reporting obligation, the institution remains responsible for compliance.

19. Case Law 6 — Aptiva Technologies FZE v Liberty Steel Group Holdings [2024] DIFC CFI 076

This case concerned a software supply agreement.

The Court dealt with:

• software licensing;
• contractual obligations;
• fitness for purpose;
• expert evidence;
• loss;
• mitigation.

The Court noted that expert evidence would have been required to establish a technical fitness-for-purpose allegation, but no such evidence had been produced.

Machine-learning governance significance

This illustrates the principle:

Technical assertions require appropriate technical evidence.

An AI system should not manufacture or substitute for missing expert evidence.

For example:

“The model says the software was defective.”

That is not automatically sufficient.

The court still needs reliable evidence establishing the relevant technical proposition.

20. Case Law 7 — Gate Mena DMCC v Tabarak Investment Capital Ltd [2023] DIFC CA 002

The Court of Appeal considered the earlier cryptocurrency litigation and ordered a retrial concerning a particular issue, remitting the case to the Digital Economy Court.

Governance significance

This case illustrates an important principle of algorithmic accountability:

Legal conclusions remain reviewable.

If a technological or judicial process produces an incorrect legal conclusion, there must be mechanisms for:

• appeal;
• reconsideration;
• retrial;
• correction.

A machine-learning system should therefore never be designed as an irreversible decision-maker.

21. Case Law 8 — Alarabi Investments Ltd v Cron AI Ltd [2026] DIFC CFI 030/2025

This is a particularly relevant recent case because the defendant itself was an AI company.

The DIFC Court dealt with an application concerning a default judgment and an attempted application to set it aside. The case illustrates that disputes involving AI businesses remain subject to ordinary procedural rules.

Governance significance

The fact that a company operates in AI does not create a separate procedural universe.

AI companies remain subject to ordinary legal rights, obligations and procedural safeguards.

22. What These Cases Demonstrate

These cases do not establish a single UAE doctrine specifically titled "machine-learning governance of legal systems."

Instead, they collectively provide principles relevant to governance:

23. AI Evidence vs AI Decision-Making

This distinction is essential.

AI as evidence

AI analyses information and gives the court an output.

Example:

"The probability that these transactions were generated by the same actor is 94%."

The court evaluates that output.

AI as decision-maker

The system itself determines:

"Liability established."

The second model raises substantially greater governance concerns.

Therefore:

AI-assisted adjudication is not necessarily the same as AI adjudication.

24. Machine Learning and Judicial Discretion

Civil law inevitably requires some judicial assessment.

Examples include:

• good faith;
• reasonableness;
• causation;
• proportionality;
• moral damage;
• mitigation;
• contractual interpretation;
• credibility.

These concepts cannot always be reduced to numerical probabilities.

A machine-learning system may identify relevant factors.

But the court must determine their legal weight.

25. Predictive Justice

Machine learning may be used to predict:

• likely case outcomes;
• likely damages;
• likely procedural developments;
• likely settlement values.

This creates a major governance distinction:

Prediction is not adjudication.

A model might predict:

“Similar cases succeeded 75% of the time.”

That does not mean:

“This claimant legally wins.”

Each dispute remains dependent upon its:

• applicable law;
• facts;
• evidence;
• procedural history.

26. Machine Learning and Precedent

A machine-learning system may identify hundreds of similar cases.

But similarity is not the same as legal authority.

The system must distinguish:

Binding authority

Applicable law or binding precedent where relevant.

Persuasive authority

Useful but not necessarily binding.

Factually similar decision

Similarity alone does not determine legal force.

Outdated decision

Historical cases may reflect superseded legislation.

This is particularly important in the UAE because legislative reforms can change the applicable legal framework.

27. Data Drift and Legal Drift

Machine-learning governance faces two different types of change.

Data drift

The underlying data changes.

Legal drift

The law changes.

For example:

Model trained on 2018–2024 cases

↓

New legislation takes effect in 2026

↓

Historical patterns may no longer accurately represent current law.

Therefore:

Legal AI must be continuously updated for legislative change.

A model that predicts legal outcomes from outdated cases can become systematically unreliable.

28. The New Civil Transactions Law

The UAE's new Civil Transactions Law, Federal Decree by Law No. 25 of 2025, became effective on 1 June 2026.

This creates a particularly important governance issue.

A machine-learning model trained primarily on judgments under the former Civil Transactions Law may continue identifying old legal patterns.

Therefore, governance should require:

• legislative-version tracking;
• effective-date awareness;
• jurisdiction identification;
• historical-law labelling;
• model updating;
• human verification.

29. Jurisdictional Governance

The UAE contains different legal environments.

Mainland UAE

Primarily federal/local civil-law institutions.

DIFC

Common-law-influenced English-language jurisdiction with its own legislation and courts.

The DIFC Courts themselves describe their jurisdiction as distinct from the UAE's Arab-language civil-law system.

ADGM

Another separate common-law-based financial jurisdiction.

Therefore:

A machine-learning model must know which legal system applies.

A model trained on DIFC cases cannot automatically treat those cases as Federal Supreme Court precedent.

30. Model Governance Architecture

A UAE legal ML system should ideally have at least eight layers.

Layer 1 — Legal authority

Identify applicable:

• legislation;
• regulations;
• binding decisions;
• contractual provisions.

Layer 2 — Jurisdiction

Identify:

• mainland;
• DIFC;
• ADGM;
• other applicable jurisdiction.

Layer 3 — Data

Verify:

• authenticity;
• completeness;
• currency;
• provenance.

Layer 4 — Model

Test:

• accuracy;
• bias;
• robustness;
• reliability.

Layer 5 — Explanation

Record:

• important factors;
• data sources;
• model limitations.

Layer 6 — Human review

A qualified human evaluates the output.

Layer 7 — Audit

Maintain logs showing:

• input;
• output;
• version;
• date;
• reviewer.

Layer 8 — Appeal/correction

Permit challenge and correction.

31. Auditability

Every significant AI-assisted legal decision should ideally leave an audit trail.

For example:

Model Version 4.2
Data cut-off: 1 June 2026
Relevant legislation identified: X
Cases identified: Y
Risk flag: Z
Human reviewer: Judicial Officer
Final decision: Human judicial determination

This allows later investigation of:

• error;
• bias;
• outdated data;
• system malfunction;
• improper use.

32. Accountability

The most important governance question is:

Who is legally responsible when the algorithm is wrong?

Possible actors include:

• software developer;
• court administrator;
• vendor;
• data provider;
• legal practitioner;
• regulator;
• institution;
• human decision-maker.

A governance framework should avoid the "algorithm made me do it" problem.

An institution should not escape legal responsibility merely because a machine produced the initial output.

33. Cybersecurity

Legal AI systems contain extremely sensitive information.

A breach could expose:

• pleadings;
• evidence;
• trade secrets;
• personal data;
• financial information;
• privileged communications.

The Graciela case demonstrates the potentially severe consequences of unauthorised interference with IT systems.

The Aegis case similarly demonstrates the legal consequences of compromised digital communications.

Therefore:

Cybersecurity is part of legal-AI governance, not merely an IT issue.

34. Confidentiality

The DIFC AI guidance expressly warns lawyers against entering confidential client information into AI systems without appropriate safeguards and requires consideration of applicable data-protection law.

A court-related ML system should therefore have:

• access controls;
• encryption;
• data segregation;
• confidentiality protocols;
• retention controls;
• logging;
• vendor controls.

35. Human-in-the-Loop Model

A useful UAE model is:

Machine learning

↓

Recommendation

↓

Human legal review

↓

Reasoned judicial decision

↓

Appeal/review

This is preferable, from a governance perspective, to:

Data

↓

Algorithm

↓

Automatic judgment

36. Role of the Digital Economy Court

The DIFC Digital Economy Court is particularly significant.

Its jurisdiction expressly covers:

• artificial intelligence;
• blockchain;
• digital assets;
• databases;
• cloud data;
• digital payment platforms;
• fintech;
• Web3-related technologies. 

Its rules also permit:

AI-driven smart forms and decision-tree software.

The Court has also announced specialised digital-custodian and blockchain-intelligence capabilities through third-party service providers for appropriate complex cases.

This shows that UAE judicial technology governance is moving beyond simple electronic filing toward technology-assisted judicial infrastructure.

37. But Smart Forms Are Not Autonomous Judges

The distinction is important.

An AI-driven form may:

• ask relevant questions;
• organise information;
• identify missing fields;
• classify a claim;
• guide users through procedural steps.

That is different from:

The AI independently decides the legal rights of the parties.

Part 58's smart-form mechanism is therefore better understood as a procedural and administrative technology, unless and until a separate legal rule provides otherwise.

38. Machine Learning and Access to Justice

Properly governed ML can improve:

• access to legal information;
• document organisation;
• translation;
• procedural guidance;
• case management;
• identification of relevant evidence;
• processing speed.

But poor governance can produce:

• automated exclusion;
• unexplained decisions;
• discriminatory outcomes;
• incorrect legal information;
• privacy violations.

Thus:

Efficiency must be balanced with legality and procedural fairness.

39. Machine Learning and Expert Evidence

Machine-learning output may itself become a subject of expert evidence.

An expert may need to explain:

• training methodology;
• error rate;
• model architecture;
• validation;
• data quality;
• bias;
• reproducibility.

The Graciela case is useful by analogy because the Court carefully considered technical forensic evidence while retaining the judicial function of deciding the ultimate legal issue.

40. Machine Learning and Evidence Law

A legal ML system should distinguish:

Primary evidence

Original digital record.

Metadata

Information about creation, transmission or modification.

Expert interpretation

Technical explanation.

AI-generated inference

A model's conclusion based on available data.

These categories should not automatically receive identical evidentiary weight.

41. The Black-Box Problem

Suppose:

Algorithm → "High fraud probability: 92%."

The court should ask:

1. What data was used?
2. Was the data accurate?
3. What variables were considered?
4. What model was used?
5. What is its known error rate?
6. Was the model validated?
7. Was the model applicable to this jurisdiction?
8. Could the output be independently reproduced?
9. Was human review conducted?

Without answers, the numerical output may have limited legal value.

42. Governance of Automated Dispute Resolution

DIFC Part 58 expressly includes claims involving automatic dispute-resolution processes within its Digital Economy Court framework.

Automated dispute resolution can be useful for:

• small-value claims;
• standardised disputes;
• straightforward contractual issues.

But governance should provide:

• human escalation;
• error correction;
• appeal/review;
• disclosure;
• evidence submission;
• jurisdictional checks.

43. Liability for Defective Legal AI

Suppose a legal-AI provider supplies a model that incorrectly identifies the applicable UAE law.

Potential issues could include:

• contractual breach;
• professional negligence;
• data-protection violations;
• misrepresentation;
• professional responsibility;
• consequential damages.

The Aptiva case demonstrates the importance of contractual and technical analysis when software performance is disputed.

44. Governance Principles

The UAE machine-learning legal framework can be organised around 10 principles:

1. Legality

AI must operate within applicable law.

2. Human authority

Human judicial authority remains central.

3. Transparency

Relevant AI use should be disclosed.

4. Accuracy

AI outputs require verification.

5. Explainability

Material decisions should be capable of meaningful explanation.

6. Accountability

A responsible human or institution must be identifiable.

7. Data protection

Personal and confidential information must be protected.

8. Bias control

Models should be tested for discriminatory or distorted outcomes.

9. Auditability

Significant AI activity should be traceable.

10. Contestability

Affected parties should have appropriate mechanisms to challenge erroneous outcomes.

45. Practical Example

Imagine a UAE court receives 50,000 documents in a commercial dispute.

An ML system classifies them into:

• relevant;
• potentially relevant;
• irrelevant.

The system then identifies 500 documents for judicial review.

Proper governance

Step 1: AI classifies.

Step 2: Lawyers review.

Step 3: Parties challenge classifications where necessary.

Step 4: Court resolves disputed relevance.

Step 5: Judge determines facts.

Step 6: Judge applies law.

Improper governance

AI classifies 500 documents → all other evidence automatically excluded → judgment issued.

The second approach risks converting an evidentiary tool into an adjudicative authority without adequate procedural safeguards.

46. UAE Governance Formula

For examination purposes:

Machine Learning Governance = Data Governance + Legal Authority + Human Oversight + Transparency + Explainability + Accountability + Security + Review

And:

AI Output ≠ Legal Rule

Prediction ≠ Proof

Classification ≠ Liability

Automation ≠ Judicial Authority

47. Important Case-Law Revision Table

Important: Most of these are DIFC authorities, not Federal Supreme Court precedents. Their value for mainland UAE civil law is principally comparative and illustrative unless the relevant DIFC law applies.

48. Key Challenges for UAE Legal-System ML

A. Algorithmic bias

Historical data can reproduce historical patterns.

B. Black-box decisions

The parties may not understand why an output was produced.

C. Outdated law

Models may rely on superseded legislation.

D. Hallucination

Generative AI may invent cases or statutory provisions.

E. Data leakage

Confidential legal information may be exposed.

F. Cyberattack

Legal databases are attractive targets.

G. Automation bias

Humans may accept machine recommendations too readily.

H. Accountability gap

It may be unclear who is responsible for an incorrect output.

I. Jurisdictional confusion

DIFC, ADGM and mainland UAE rules may be incorrectly mixed.

J. Excessive standardisation

Individual factual circumstances may be overlooked.

49. Recommended Governance Architecture for UAE Courts

A robust system could follow:

Level 1 — Legal-source validation

Every AI-generated legal proposition should be checked against:

• legislation;
• official judgments;
• procedural rules.

Level 2 — Jurisdiction filter

Identify:

• Federal UAE;
• Dubai;
• Abu Dhabi;
• DIFC;
• ADGM.

Level 3 — Temporal filter

Determine the law applicable on the relevant date.

Level 4 — Evidence validation

Verify source authenticity.

Level 5 — Model validation

Test accuracy and bias.

Level 6 — Human review

Require qualified human assessment.

Level 7 — Reasoned output

Explain how the result was reached.

Level 8 — Audit trail

Maintain records.

Level 9 — Challenge

Permit appropriate review.

Level 10 — Continuous monitoring

Update the system when legislation and jurisprudence change.

50. Conclusion

Machine learning governance of legal systems in the UAE is best understood not as a question of whether courts should use AI, but as a question of how technological assistance can coexist with legal authority, evidence, fairness and human judicial responsibility.

The UAE's DIFC framework is particularly advanced. The Digital Economy Court expressly covers AI and other emerging technologies, while Part 58 permits AI-driven smart forms and digital proceedings. The DIFC Courts' 2023 AI guidance further establishes practical principles of transparency, accuracy, verification, confidentiality, data protection and avoidance of over-reliance on AI.

The case law illustrates the same broader direction. Graciela demonstrates that sophisticated technical evidence remains subject to judicial evaluation; Aegis Resources shows that technology-related liability is fact-specific; Gate Mena demonstrates adaptation of legal reasoning to digital assets; Al Ramz Capital illustrates continuing institutional responsibility in technologically sophisticated financial regulation; and Aptiva Technologies shows why technical assertions require appropriate evidence.

The central principle for examination is:

“Machine learning may assist the UAE legal system in processing information, identifying patterns and managing disputes, but it should not become an independent source of law or an unreviewable substitute for human judicial reasoning.”

In short:

Data can inform justice; algorithms can assist justice; but legal authority must remain grounded in law, evidence, procedural fairness and accountable judicial decision-making.