Civil Law and UAE — Lack of Unified Legal Epistemology in AI Law Systems

1. Introduction

The expression “lack of unified legal epistemology” refers to the absence of a single, consistent method for determining:

• what counts as legally reliable knowledge;
• how facts generated or analysed by AI should be proved;
• how algorithmic conclusions should be interpreted;
• how much weight should be given to automated recommendations;
• who is responsible when an AI system produces an error;
• how a court should reconcile algorithmic output with human testimony, expert evidence and legal reasoning.

This issue is particularly important in the UAE because the country does not presently have one comprehensive federal AI statute governing all civil, judicial and algorithmic applications. Instead, AI is addressed through a combination of civil law, evidence law, data-protection legislation, sector-specific regulation, emirate-level rules, free-zone legislation and non-binding AI governance instruments.

The problem is therefore not simply that “AI law is incomplete.” The deeper issue is that different legal systems may use different methods for deciding what constitutes trustworthy knowledge.

2. Meaning of Legal Epistemology

Legal epistemology

Legal epistemology concerns the question:

How does the legal system know that something is true, reliable or legally relevant?

Traditional civil litigation generally relies upon:

1. documentary evidence;
2. witness testimony;
3. expert evidence;
4. admissions;
5. presumptions;
6. physical evidence;
7. judicial reasoning;
8. statutory interpretation.

AI introduces additional sources:

• machine-generated predictions;
• automated classifications;
• algorithmic risk scores;
• facial-recognition outputs;
• synthetic documents;
• generative-AI legal research;
• automated valuations;
• predictive analytics;
• AI-generated evidence.

The fundamental question becomes:

Can an algorithmic conclusion be treated as knowledge in the same way as human-generated evidence?

The UAE currently does not have one unified answer applicable across every civil-law context.

3. Why the UAE Has an Epistemological Fragmentation Problem

The UAE's legal structure itself contributes to the issue.

There are:

A. Federal civil-law institutions

Including:

• federal legislation;
• federal courts;
• UAE Evidence Law;
• Civil Transactions Law;
• Personal Data Protection Law.

B. Emirate-level legal systems

Dubai and Abu Dhabi, for example, have substantial local legislation and judicial institutions.

C. DIFC

The DIFC operates a separate common-law-based legal system and courts.

D. ADGM

ADGM also has its own common-law-based legal system.

E. Sector regulators

AI may additionally interact with:

• financial regulation;
• healthcare;
• telecommunications;
• data protection;
• cybersecurity;
• digital assets;
• employment;
• consumer protection.

The result is legal pluralism rather than a single AI epistemological model. Current sources describe the UAE AI framework as a layered system rather than a comprehensive standalone AI statute.

4. The Central Epistemological Conflict

Traditional civil law asks:

What evidence proves the fact?

AI systems often operate differently:

What probability does the model assign to the fact?

These are not necessarily the same question.

For example, an AI system may state:

“There is an 87% probability that this transaction is fraudulent.”

But a court must decide:

• What data produced the 87%?
• Was the data accurate?
• Was it complete?
• Was it biased?
• Was the model appropriately trained?
• Can the opposing party challenge the methodology?
• Can the expert reproduce the result?
• Was the model changed after the transaction?
• Does probability establish the legal elements of fraud?

Therefore:

Algorithmic probability ≠ legal proof automatically.

5. Traditional Legal Epistemology vs AI Epistemology

This difference explains why AI creates an epistemological problem, not merely a technological problem.

6. UAE's Emerging Human-Oversight Principle

The UAE AI Charter expressly emphasises:

• transparency;
• privacy;
• human oversight;
• accountability;
• responsible AI;
• correcting errors and biases.

The Charter describes human judgment and oversight as having an irreplaceable role.

This provides an important conceptual foundation:

AI can assist legal decision-making, but technological output does not automatically replace human legal judgment.

However, the Charter is not equivalent to a comprehensive AI statute establishing detailed evidentiary rules for every AI system.

7. DIFC's More Developed AI Procedural Approach

The DIFC provides one of the clearest UAE examples of an emerging AI epistemology.

The DIFC Courts' Practical Guidance Note No. 2 of 2023 requires transparency regarding AI-generated content and emphasises:

• verification;
• reliability;
• disclosure;
• consideration of training data;
• algorithmic limitations;
• possible bias;
• confidentiality;
• human oversight;
• avoidance of excessive reliance on generative AI. 

This is significant because it creates a relatively clear epistemological rule:

AI output is potentially useful, but its reliability must be independently assessed.

8. Case Law 1 — Klesta Eshja v Salah Masri

Klesta Eshja & Hair Creators Salon LLC v Salah Masri & Others

[2024] DIFC CFI 066

This is one of the most directly relevant UAE cases concerning generative AI.

The defendants' amended pleadings were prepared substantially with AI assistance and contained false references and misleading material. The DIFC Court ordered the amended defences struck out and subsequently dealt with the resulting costs. In March 2026, the court assessed costs at AED 343,953.75.

Principle

AI assistance does not transfer responsibility for legal submissions from the lawyer or litigant to the AI system.

Epistemological significance

The court effectively insisted upon:

AI output → human verification → legally responsible submission.

The case therefore demonstrates that machine-generated legal content does not become authoritative merely because it appears legally sophisticated.

9. Case Law 2 — Arabyads Holding Ltd v Gulrez Alam Marghoob Alam

Arabyads Holding Ltd v Gulrez Alam Marghoob Alam

[2025] ADGMCFI 0032

This ADGM decision involved fictitious or non-existent legal authorities appearing in litigation material. The court imposed approximately AED 282,508 in wasted costs against the relevant representatives for failures associated with verification of the authorities. Contemporary UAE AI-law research identifies the case as an important example of judicial treatment of AI-generated legal material.

Principle

A lawyer's professional duty to verify legal authorities cannot be delegated to an AI system.

Epistemological significance

The case distinguishes:

information generated by AI

from

legally verified knowledge.

That distinction is central to legal epistemology.

10. Case Law 3 — Alarabi Investments Ltd v Cron AI Ltd

Alarabi Investments Limited v Cron AI Ltd

[2025] DIFC CFI 030

The defendant's corporate identity included an AI-related business, but the DIFC Court dealt with the dispute through ordinary procedural mechanisms, including a default judgment and an application concerning its setting aside. The matter demonstrates that the technological nature of a company does not itself remove it from ordinary legal principles.

Principle

An AI-related enterprise remains subject to:

• jurisdiction;
• procedural rules;
• pleadings;
• judgments;
• costs;
• enforcement.

Epistemological significance

Technology does not create an automatic alternative system of legal truth.

The court remains the institution determining:

what facts are legally established and what consequences follow.

11. Case Law 4 — Techteryx Ltd v Aria Commodities DMCC

Techteryx Ltd v Aria Commodities DMCC & Others

[2025] DIFC DEC 001

This case was dealt with by the DIFC Digital Economy Court and involved sophisticated digital-asset issues, including proprietary injunctions and tracing of assets.

The case demonstrates the development of specialised judicial capacity for technologically complex disputes.

Principle

Technological complexity can justify specialised procedural expertise without necessarily creating a completely separate theory of legal responsibility.

Epistemological significance

The Digital Economy Court represents an institutional response to technological complexity:

complex technology → specialised judicial competence → ordinary principles of proof and adjudication remain relevant.

12. Case Law 5 — Gate MENA DMCC v Tabarak Investment Capital Ltd

Gate MENA DMCC v Tabarak Investment Capital Ltd

[2024] DIFC DEC 002

The case was heard in the DIFC Digital Economy Court and concerned a sophisticated digital-economy dispute.

Its importance for AI epistemology is institutional rather than purely doctrinal: the DIFC has created a specialised forum capable of handling technologically complex disputes rather than requiring courts to treat every emerging technology through identical procedural assumptions.

Principle

Legal institutions can specialise in technological disputes while preserving judicial control over legal questions.

Epistemological significance

This supports a human-expert model rather than an AI-autonomous model.

13. Case Law 6 — Ganesan Muthiah v Abdul Rahman Mohammad

Ganesan Muthiah v Abdul Rahman Mohammad

[2025] DIFC CFI 055

The dispute involved jurisdictional conflict between the DIFC Courts and Dubai Courts. The Conflict of Jurisdiction Tribunal determined that the Dubai Courts had jurisdiction, following which the DIFC proceedings were dismissed.

Principle

A technological or commercial dispute does not eliminate the need to identify the legally competent judicial system.

Epistemological significance

This is important because legal knowledge is jurisdiction-dependent.

An AI system trained predominantly on:

• English common law;
• DIFC cases;
• foreign cases;

could generate an apparently sophisticated answer that is nevertheless wrong if the applicable law is mainland UAE civil law.

Thus:

Correct reasoning under the wrong legal system can still produce an incorrect legal result.

14. Case Law 7 — Naatiq v Nabeeh

Naatiq v Nabeeh

[2024] DIFC ARB 018

This authority is relevant to the broader problem of jurisdictional and procedural fragmentation.

Where competing jurisdictions or proceedings exist, legal decision-making must first determine which legal framework controls.

Epistemological significance

AI systems frequently produce answers by aggregating multiple legal sources.

But aggregation is not the same as legal selection.

The legal system must determine:

1. which jurisdiction applies;
2. which law applies;
3. which court has authority;
4. which evidentiary standard applies;
5. which precedent has legal relevance.

Therefore:

More legal information ≠ better legal knowledge.

15. The Evidence Problem

AI creates several evidentiary difficulties.

A. Authenticity

Who created the document?

B. Integrity

Has the document been altered?

C. Provenance

Where did the information originate?

D. Reliability

Was the AI system accurate?

E. Reproducibility

Can another expert reproduce the result?

F. Explainability

Can the court understand why the AI reached its conclusion?

G. Bias

Was the training data or algorithm systematically biased?

H. Accountability

Who is legally responsible for the result?

These questions are particularly important because UAE evidence law operates around established concepts of proof, presumptions and judicial evaluation, while AI can produce evidence whose internal reasoning is technically difficult to reconstruct.

16. The “Black Box” Problem

A major epistemological problem is the black-box algorithm.

Suppose an AI system rejects a loan application.

The affected person asks:

“Why?”

The operator responds:

“The model determined that your risk score was too high.”

That answer may not provide a legally meaningful explanation.

A court may instead need:

• input data;
• decision criteria;
• model methodology;
• relevant variables;
• error rates;
• training information;
• audit logs;
• human intervention;
• explanation of the decision.

The legal problem therefore becomes:

Can a person effectively challenge a decision without understanding the basis on which the algorithm produced it?

17. Data Protection and AI Epistemology

The UAE Personal Data Protection Law is particularly important because it expressly recognises automated processing. It defines automated processing as processing performed through an automated system either independently or with limited human supervision/intervention.

This creates an important connection:

Data → AI processing → inference → decision → legal consequence.

If the underlying data is:

• inaccurate;
• unlawfully obtained;
• excessive;
• improperly processed;

then the reliability of the resulting AI inference may also be questioned.

Therefore:

Data governance is part of legal epistemology.

18. AI Bias as an Epistemological Problem

Bias is often discussed as an equality problem.

But it is also an epistemological problem.

Suppose an algorithm was trained using historically biased data.

The system may identify the historical pattern as a “fact”.

But:

historical frequency ≠ legal correctness.

For example:

If historical decisions disproportionately rejected a particular category of applicants, an AI system trained on those decisions might reproduce the pattern.

The algorithm may be statistically consistent while still producing legally problematic results.

Thus:

Statistical accuracy and legal validity are different concepts.

19. AI Hallucination and Legal Knowledge

Generative AI can produce:

• nonexistent cases;
• incorrect statutory provisions;
• fabricated quotations;
• incorrect case citations;
• outdated legal rules.

The DIFC guidance expressly warns practitioners about misleading or incorrect information and requires verification of AI-generated content.

The Klesta and Arabyads matters demonstrate why this is important.

Legal principle

AI-generated content has no independent evidentiary authority merely because it was produced by a sophisticated model.

Its legal value depends on verification.

20. Lack of Unified Epistemology in UAE AI Law

The problem can be represented as follows:

Federal UAE

Primarily civil-law methodology.

↓

DIFC

Common-law methodology + specialised AI procedural guidance.

↓

ADGM

Common-law methodology + professional accountability through judicial decisions.

↓

Sector regulators

Specialised regulatory knowledge.

↓

AI systems

Statistical/probabilistic knowledge.

These systems do not necessarily answer the same question in the same way.

21. Why This Matters for Civil Liability

Traditional civil liability normally asks:

1. Was there a wrongful act?

2. Was there damage?

3. Is there causation?

4. Who is responsible?

AI complicates every stage.

Suppose an autonomous system causes financial loss.

Possible responsible actors include:

• developer;
• manufacturer;
• deployer;
• operator;
• data provider;
• owner;
• professional who relied upon the system.

The AI itself is not ordinarily treated as a human legal actor simply because it acts autonomously.

Therefore, existing civil-liability principles must determine how responsibility is allocated among human and corporate actors.

22. AI and Causation

Causation becomes especially difficult where multiple AI components interact.

Example:

Developer error

↓

bad training data

↓

algorithmic output

↓

human reliance

↓

financial decision

↓

loss

Which event legally caused the damage?

Traditional civil law tends to search for a legally attributable causal chain.

AI can instead create distributed causation.

This is one of the clearest areas where existing civil-law concepts may require careful adaptation rather than simply applying the language of “the algorithm made the decision.”

23. AI and the Principle of Human Responsibility

The UAE AI Charter expressly emphasises human oversight and accountability.

The DIFC AI guidance similarly states that AI should assist rather than replace the integral human decision-making required in preparing evidence and submissions.

This produces an important principle:

Automation of reasoning does not automatically mean automation of legal responsibility.

The human or legal entity deploying AI remains potentially responsible for lawful use of the technology.

24. AI and Judicial Decision-Making

The most difficult epistemological question is:

Can AI determine a civil dispute?

There is an important difference between:

AI-assisted adjudication

AI helps:

• search authorities;
• classify documents;
• identify patterns;
• organise evidence;
• detect inconsistencies.

and:

AI autonomous adjudication

AI effectively determines:

• facts;
• credibility;
• liability;
• damages;
• legal interpretation;
• final judgment.

The UAE framework currently provides substantially stronger support for the assistance/oversight model than for treating AI as an autonomous judicial decision-maker. The UAE AI Charter stresses human oversight, while the DIFC's litigation guidance expressly rejects replacing integral human decision-making with AI.

25. Digital Economy Court and Epistemological Specialisation

The DIFC's Digital Economy Court is significant because its rules expressly cover disputes involving:

• fintech;
• digital assets;
• distributed ledger technology;
• substantial or complex databases;
• artificial intelligence and AI-dependent devices or components. 

This is an institutional response to the epistemological complexity of digital disputes.

Instead of assuming that ordinary judicial expertise is always sufficient, the system creates a specialist forum.

But the court remains a human judicial institution.

Therefore:

Technological complexity → specialised legal expertise, not necessarily algorithmic adjudication.

26. The Core Problem: Multiple Standards of “Truth”

AI systems may recognise:

Statistical truth

“What happens most frequently?”

Predictive truth

“What is most likely to happen?”

Data truth

“What does the dataset indicate?”

Legal truth

“What fact has been proved according to the applicable legal rules?”

Judicial truth

“What conclusion does the competent court reach after evaluating admissible evidence?”

These concepts overlap, but they are not identical.

This is the heart of the UAE's AI epistemological challenge.

27. Lack of Unified Explainability Standard

Another major difficulty is the absence of one comprehensive UAE-wide standard specifying:

• what an AI explanation must contain;
• when algorithmic disclosure is mandatory;
• what level of technical transparency is required;
• how proprietary algorithms should be treated;
• how parties obtain model documentation;
• how courts evaluate algorithmic confidence;
• how AI-generated evidence should be challenged.

The DIFC has specific practitioner guidance, but current sources indicate that the UAE still lacks a comprehensive AI statute covering the entire country.

28. Consequences of Fragmented Legal Epistemology

A. Inconsistent outcomes

Different jurisdictions may approach identical AI evidence differently.

B. Forum shopping

Parties may prefer a jurisdiction with clearer technology rules.

C. Evidentiary uncertainty

Parties may not know what AI-generated material will be accepted.

D. Increased expert costs

Technical experts may be required to explain AI systems.

E. Accountability gaps

It may be unclear whether responsibility lies with the developer or user.

F. Procedural inequality

A party with greater technological resources may have greater capacity to challenge an algorithm.

G. Privacy risks

AI investigation may require processing large quantities of personal information.

H. Reduced legal certainty

Different standards may make legal outcomes harder to predict.

29. AI Evidence — Proposed UAE Civil-Law Analytical Framework

Until a comprehensive unified regime develops, an AI-related civil dispute can be analysed through the following sequence:

Step 1 — Identify jurisdiction

Mainland UAE, DIFC, ADGM or another applicable system?

Step 2 — Identify applicable law

Civil law, commercial law, data law, IP law, financial regulation, etc.

Step 3 — Identify the AI function

Is AI:

• evidence?
• decision-maker?
• analytical tool?
• contractual service?
• product?
• autonomous system?

Step 4 — Establish provenance

Where did the AI output originate?

Step 5 — Establish reliability

Was the output independently verified?

Step 6 — Examine explainability

Can the methodology be sufficiently understood and challenged?

Step 7 — Examine data legality

Was personal or confidential data lawfully processed?

Step 8 — Determine human responsibility

Who selected, deployed or relied upon the system?

Step 9 — Apply ordinary civil-law principles

• fault;
• damage;
• causation;
• contractual obligation;
• good faith;
• proportionality;
• compensation.

Step 10 — Apply specialised AI rules/guidance

Where applicable.

30. Relationship With the New UAE Civil Transactions Law

The new Civil Transactions Law provides the general private-law architecture within which many AI disputes will have to be analysed.

For example, its contractual framework requires attention to:

• consent;
• intention;
• surrounding circumstances;
• commercial custom;
• justice;
• good faith.

Article 120 expressly provides that a contract is to be interpreted in a manner achieving justice and good faith between the parties.

This is important for AI contracts such as:

• AI-as-a-service agreements;
• algorithm licensing;
• automated decision contracts;
• cloud-AI agreements;
• machine-learning development agreements;
• AI consultancy contracts.

31. AI Contracts and Epistemological Risk

An AI contract should ideally identify:

1. what the AI system is expected to do;
2. what data will be supplied;
3. who owns the resulting data;
4. acceptable accuracy levels;
5. explainability requirements;
6. audit rights;
7. human review;
8. cybersecurity obligations;
9. confidentiality;
10. liability for erroneous outputs;
11. regulatory compliance;
12. termination rights.

Otherwise, a dispute may arise over whether the AI system delivered the contractually promised result.

32. Important Distinction: AI Is Not Automatically a Legal Person

One possible misconception is:

“The AI made the decision, therefore the AI is liable.”

That does not follow automatically.

Civil liability normally requires an identifiable legal basis connecting:

conduct → damage → causation → legally responsible person/entity.

AI may be the mechanism through which harm occurred without itself becoming the legal bearer of liability.

This makes the allocation of responsibility among developers, deployers and users a central future issue.

33. Six-Case Revision Table

The first two are the most directly AI-specific. The remaining authorities are better understood as digital/technological and institutional analogies, rather than cases establishing a general UAE doctrine of AI liability. That distinction is important because reported mainland UAE case law specifically deciding AI civil-law questions remains limited.

34. Critical Legal Analysis

The lack of unified legal epistemology does not necessarily mean that UAE law is incapable of dealing with AI.

Existing civil-law principles can already answer many questions.

For example:

AI causes contractual breach

→ apply contract law.

AI output causes measurable damage

→ examine civil liability.

AI processes personal data

→ examine data-protection law.

AI-generated document is submitted to court

→ examine evidence and procedural rules.

AI system produces false legal authorities

→ examine professional responsibility and procedural sanctions.

The difficulty arises when AI produces new forms of evidence or causation that do not fit neatly within traditional categories.

35. Future Direction

A more unified UAE AI legal epistemology would ideally establish common principles for:

1. Human oversight

A human remains responsible for consequential legal decisions.

2. Explainability

Affected parties receive an appropriate explanation of significant automated decisions.

3. Auditability

AI systems can be independently examined when legally necessary.

4. Provenance

The origin and processing history of AI-generated evidence can be established.

5. Reliability

AI output must be independently validated before receiving substantial legal weight.

6. Contestability

Affected parties must have meaningful opportunities to challenge AI conclusions.

7. Accountability

Responsibility should be allocated among developers, deployers and users.

8. Data legality

AI decisions should be traceable to lawfully processed information.

9. Jurisdictional consistency

Federal, emirate-level and free-zone systems should minimise contradictory approaches.

10. Human judicial authority

AI should support rather than replace the fundamental judicial responsibility to determine legal rights.

36. Conclusion

The lack of unified legal epistemology in UAE AI law systems is ultimately a problem of how law understands knowledge.

Traditional civil law relies heavily upon:

evidence + human evaluation + legal rules + judicial reasoning.

AI introduces:

data + statistical inference + automated prediction + machine-generated reasoning.

The two systems are not automatically interchangeable.

The UAE has begun addressing this gap through the AI Charter, data-protection rules, DIFC AI guidance, the Digital Economy Court and emerging judicial decisions concerning AI-generated legal material.

The most important emerging principle is therefore:

AI may generate information, predictions or legal content, but legal validity depends upon human verification, applicable law, evidentiary reliability, jurisdiction and judicial assessment.

Short exam formula

AI Output ≠ Legal Truth

Data ≠ Evidence

Probability ≠ Proof

Prediction ≠ Judgment

Automation ≠ Responsibility

Algorithmic Reasoning + Human Verification + Legal Rules = Legally Usable AI Knowledge

And the broader UAE problem can be summarised as:

“The challenge is not merely regulating AI; it is creating a coherent legal method for deciding when AI-generated information should count as legally reliable knowledge.” 