Civil Law and UAE: Machine-Negotiated Settlements Without Human Input

1. Meaning

A machine-negotiated settlement without human input is a settlement process in which software, an AI system, algorithm, or automated platform independently:

1. analyses the parties' claims and counterclaims;
2. estimates possible outcomes;
3. proposes settlement amounts or terms;
4. makes or receives offers;
5. adjusts its offers according to predetermined rules or machine-learning outputs; and
6. ultimately accepts or concludes a settlement without a human reviewing or approving the final agreement.

Example:

Company A claims AED 1 million from Company B. An AI platform analyses the contract, evidence and historical settlement data. It offers AED 650,000, automatically responds to B's counteroffer, and finally accepts AED 720,000. No director, lawyer, mediator or party personally approves the final settlement.

The central UAE legal question is therefore:

Can an automated system create a legally binding settlement when no human directly makes the final decision?

The answer requires separating electronic formation of contracts from legal authority to compromise rights.

2. Current UAE Legal Framework

The issue is particularly important because the UAE has expressly recognised automated electronic contracting.

Under Federal Decree-Law No. 46 of 2021 on Electronic Transactions and Trust Services, offer and acceptance may be expressed electronically, and a contract does not lose validity merely because it is made through electronic documents. More importantly, Article 11 expressly recognises contracts formed between automated electronic mediums that have been programmed in advance for that purpose.

This provides an important foundation for algorithmic contracting.

However, a settlement agreement is not simply an ordinary automated purchase contract.

The new UAE Civil Transactions Law, Federal Decree by Law No. 25 of 2025, effective from 1 June 2026, contains specific provisions on settlement. Articles 672–676 address capacity, permissible subject matter, proof and the requirements concerning the settlement consideration.

Therefore, an automated settlement must satisfy both:

electronic/automated contract formation requirements

and

substantive requirements applicable to settlement.

3. Why Settlement Is Different From Ordinary Automated Contracting

An automated purchase is relatively straightforward:

Price + product + programmed acceptance → contract

A settlement is more complicated:

Dispute + disputed rights + compromise + authority + consideration + waiver/release → settlement

A settlement may extinguish:

• existing claims;
• future claims within the agreed scope;
• damages claims;
• contractual rights;
• counterclaims;
• guarantees;
• interest claims;
• arbitration or litigation rights.

Consequently, the legal system must ask not merely:

"Did the computer send acceptance?"

but:

"Was there legally sufficient authority to compromise these rights?"

4. Article 672 and Capacity

The new Civil Transactions Law provides that a person concluding a settlement must have the capacity to dispose, for consideration, of the rights covered by the settlement.

This creates an important distinction.

Human-controlled automation

A company authorises its director to use an AI negotiation system within:

• AED 100,000–500,000;
• specified disputes;
• specified settlement parameters.

The AI concludes a settlement within those parameters.

This is much easier to characterise legally as human-authorised automated contracting.

Completely autonomous AI

The AI itself determines:

• whether a claim should be abandoned;
• how much compensation should be accepted;
• whether a guarantee should be released;
• whether litigation should be discontinued.

Here the difficult question becomes authority and attribution.

The machine does not itself become the legal owner of the claim.

5. Machine Output Does Not Automatically Become Legal Authority

A machine-learning model may calculate:

"Expected litigation recovery = AED 610,000."

That is a statistical output.

It is not automatically:

• a legal finding;
• an admission;
• an expert determination;
• a judicial decision;
• a binding valuation.

Likewise:

"Recommended settlement = AED 550,000"

does not by itself mean that the claimant has legally agreed to settle for AED 550,000.

The legal effect depends upon the authority given to the system and the applicable rules of contract and settlement.

6. Automated Negotiation Under Article 11 of the Electronic Transactions Law

Article 11 of Federal Decree-Law No. 46 of 2021 is particularly important because it recognises contracts made between automated electronic mediums that were programmed in advance for that purpose.

This means UAE law does not necessarily require a human to physically click "accept" for every electronically generated contract.

Therefore:

No human physically pressing an acceptance button does not automatically mean that there is no contract.

However, this does not mean:

AI has unlimited legal capacity to settle any dispute.

The scope of the system's programming, authority and underlying rights remains critical.

7. The Authority Problem

Suppose a company's board authorises an AI system:

"You may settle claims up to AED 500,000."

The AI settles for AED 450,000.

There is a strong contractual argument that the automated settlement falls within the authority deliberately granted to the system.

But suppose the AI settles for AED 2 million.

The opposing party may argue:

• the system exceeded its authority;
• the company never authorised that settlement;
• the settlement should not bind the company;
• the algorithm malfunctioned;
• the settlement resulted from erroneous data;
• the AI misunderstood the dispute.

Thus, algorithmic authority should be treated as delegated authority, not independent legal personality.

8. Machine Negotiation and Agency

The most useful conceptual model is:

Principal → authorised automated system → counterparty

rather than:

AI → independent legal person → counterparty

The AI normally operates as an instrument through which the principal acts.

This is important because current UAE civil-law principles generally attach contractual rights and obligations to natural or juridical persons, rather than treating an AI model itself as an independent holder of civil rights.

Therefore:

The machine negotiates, but the legal person remains the party to the settlement.

9. The Problem of Machine Error

Suppose an AI settlement system incorrectly reads:

AED 10 million

as:

AED 1 million.

It automatically accepts AED 1 million in full settlement.

Several questions arise:

1. Was the AI authorised?
2. Was the error obvious?
3. Did the counterparty know of the error?
4. Was the system operating according to its programmed rules?
5. Was the mistake caused by faulty data?
6. Did the human principal supervise the system?
7. Does the law permit avoidance or cancellation on the particular facts?

This makes auditability extremely important.

A company using automated settlement should preserve:

• input data;
• negotiation rules;
• model version;
• prompts;
• offer history;
• acceptance event;
• authorisation limits;
• identity/authentication records;
• timestamps;
• system logs.

10. Good Faith and Automated Negotiation

A machine cannot be allowed to become a mechanism for manipulating the settlement process.

For example, an AI might be programmed to:

• conceal relevant information;
• exploit a known pricing error;
• misrepresent authority;
• generate false factual statements;
• pressure the other party through repeated automated messages.

The existence of automation does not eliminate the legal obligations governing the underlying transaction.

Therefore:

Automation changes the method of negotiation; it does not automatically eliminate substantive civil-law duties.

11. Human Input: Is It Always Legally Necessary?

There are three different models.

The third model presents the greatest uncertainty.

The strongest legal argument for enforceability is generally where the parties previously agreed that an automated system could conclude transactions within defined parameters.

12. New Civil Transactions Law and Settlement

The new Civil Transactions Law is especially relevant because it expressly regulates settlement.

The law addresses:

• capacity;
• rights capable of settlement;
• consideration;
• knowledge of the settlement subject matter;
• proof;
• settlements involving persons requiring legal assistance;
• restrictions involving personal status and public order. 

Consequently, an AI settlement engine should not be permitted to determine autonomously that every type of legal dispute is capable of settlement.

For example, an automated system should distinguish:

Potentially suitable

• commercial debt;
• invoice disputes;
• contractual damages;
• payment disputes;
• construction delay claims.

Potentially restricted or requiring special treatment

• personal-status matters;
• public-order matters;
• rights requiring statutory protection;
• claims where special approval is required;
• settlements involving persons lacking appropriate capacity.

13. Electronic Evidence

A machine-negotiated settlement will usually exist as electronic data.

Federal Decree-Law No. 46 of 2021 recognises electronic documents and provides conditions under which electronic records can satisfy requirements concerning documents and their integrity.

Therefore, an automated settlement platform should be capable of proving:

Who authorised the system?

↓

What rules were programmed?

↓

What information did the system receive?

↓

What offers were generated?

↓

What caused acceptance?

↓

Was the final settlement within the authorised limits?

This is essentially an electronic evidence chain of custody for contractual formation.

14. DIFC Position: Particularly Important

The DIFC provides an advanced UAE example because its Digital Economy Court specifically handles disputes involving:

• artificial intelligence;
• blockchain;
• big data;
• cryptocurrencies;
• cloud services;
• robotics and other emerging technologies. 

The DIFC Courts have also issued Practical Guidance Note No. 2 of 2023 concerning the use of LLMs and generative AI.

The Guidance stresses:

• transparency;
• accuracy;
• reliability;
• disclosure of AI use;
• protection of confidentiality;
• data protection;
• awareness of bias;
• verification;
• avoidance of excessive reliance on AI;
• preservation of human decision-making. 

Although this Guidance concerns AI used in proceedings rather than directly regulating autonomous settlement contracts, it is highly relevant to the governance principle:

AI should assist legal decision-making without becoming an unreviewable substitute for human legal responsibility.

15. Case Law

There is currently no well-established UAE reported case directly holding that an AI system, acting entirely without human input, has independently concluded a binding civil settlement. Therefore, the following cases are best understood as authorities concerning the underlying principles: settlement formation, authority, electronic signatures, automated systems, contractual interpretation and digital evidence.

Case 1 — Rada Trading LLC FZC v Wealth Bridge Trading & Cohenrich Energy FZE

[2021] DIFC CA 007

The dispute concerned whether communications, including emails, could potentially operate as a variation of a Settlement Agreement.

The DIFC Court of Appeal held that whether the emails constituted a valid variation depended on the evidence and applicable contractual/statutory framework.

Relevance to machine negotiation

This is important because an automated settlement system may produce:

• email offers;
• automated counteroffers;
• acceptance messages;
• electronic modifications.

The existence of an electronic communication does not by itself answer whether it legally varied a settlement.

Principle:

Electronic communication must still satisfy the legal requirements for contractual modification.

16. Case 2 — Ginette PJSC v Geary Middle East FZE & Geary Ltd

[2016] DIFC CA 005

The parties had entered into a substantial settlement agreement resolving claims and providing for payment arrangements and arbitration.

The Court considered the authority of the person signing the settlement and recognised the significance of apparent authority. It also referred to Dubai Court of Cassation Case No. 547 of 2014 concerning the authority of a company representative.

Relevance

This case demonstrates why authority is central to automated settlement.

An AI system does not need to become a legal person if it acts within authority granted by the company.

The legal question becomes:

Did the principal authorise the system to conclude this transaction?

17. Case 3 — Ondina v Olin

[2025] DIFC CFI 046

The case concerned electronic communications and whether an email exchange could satisfy a statutory signature requirement.

The Court considered the DIFC Electronic Transactions Law, including the concept of an electronic signature and attribution of an electronic signature to the person whose act it was.

The Court concluded that the relevant email communication could constitute an electronic signature because the person's name was attached to electronic information with the necessary intention.

Relevance

This supports a fundamental proposition:

Electronic form does not automatically destroy legal validity.

For machine-negotiated settlements, however, attribution remains crucial.

An automated message must be attributable to the person or entity that authorised the system.

18. Case 4 — Naho v Neukirchi

[2024] DIFC SCT 415

The case involved a Final Settlement Agreement and also considered electronic communications and electronic signatures under the DIFC Electronic Transactions Law.

Relevance

The case illustrates how courts examine:

• the actual settlement document;
• surrounding electronic communications;
• statutory signature requirements;
• the parties' conduct.

For AI settlements, this means the complete machine-generated negotiation record may become important evidence.

19. Case 5 — DIFC Investments Ltd v Dubai Islamic Bank

[2022] DIFC CFI 024

The dispute involved a construction-related settlement agreement under which a payment was made as a full and final settlement of claims and counterclaims.

Principle

A properly concluded settlement can have substantial consequences for subsequent claims.

Relevance

An AI settlement system therefore cannot treat settlement as merely another negotiation message.

Once a settlement is validly concluded:

previously existing claims may be extinguished or restricted according to its terms.

That makes the machine's authority and settlement parameters especially important.

20. Case 6 — RAK Ceramics PJSC v Assala Development SARL & Ali Chaoui

[2019] DIFC CFI 086

The case concerned a detailed Settlement Agreement containing payment obligations, security arrangements, triggering provisions and consequences of non-payment.

The Court considered the contractual consequences of failure to perform the settlement and applied UAE Civil Code principles.

Relevance

An automated settlement system must be capable of distinguishing between:

negotiation

and

final settlement with legally enforceable obligations.

A machine-generated agreement can therefore create significant downstream obligations if properly authorised and concluded.

21. Case 7 — Michael George Forbes v Robert Kidd

[2023] DIFC CFI 081

The dispute involved a substantial settlement agreement resolving multiple claims and litigation matters.

Relevance

It demonstrates the complexity that may be contained in a settlement:

• multiple claims;
• multiple parties;
• payment obligations;
• releases;
• related proceedings.

This creates a significant challenge for machine negotiation.

An algorithm may be able to calculate monetary values, but settlement often involves legal relationships extending beyond a single numerical amount.

22. Case 8 — BAM Higgs & Hill LLC v Affan Innovative Structures LLC

[2021] DIFC CFI 106

The case involved a large construction dispute and a later settlement agreement dealing with extensive claims, including delay-related claims.

Relevance

Construction settlements show why automated settlement cannot be based exclusively on monetary optimisation.

A settlement may contain:

• waiver of claims;
• defect exceptions;
• payment arrangements;
• extension-of-time consequences;
• guarantees;
• future obligations.

Thus:

A settlement is a legal allocation of rights, not merely a numerical optimisation problem.

23. Case 9 — Barclays Bank PLC v Bavaguthu Raghuram Shetty

[2020] DIFC CFI 061

The case involved evidence concerning automated systems and electronically manipulated documents. The Court considered evidence concerning automated banking processes and electronic document execution.

Relevance

This demonstrates the evidentiary problem created when automated systems are involved.

A court may need to reconstruct:

system event → automated action → data received → human awareness → legal consequence.

This is directly relevant to disputes over an AI-generated settlement.

24. Machine Negotiation and Mistake

Consider:

Claim = AED 5 million
AI settlement authority = AED 500,000–AED 1 million
AI accidentally accepts AED 100,000.

The legal analysis should include:

Step 1

Was AED 100,000 within the system's authority?

Step 2

Was the error obvious to the other party?

Step 3

Was the machine acting according to its programmed instructions?

Step 4

Did the principal create the system and authorise its operation?

Step 5

Does applicable UAE law provide a basis for avoiding or challenging the settlement?

Step 6

What does the audit trail establish?

This is much more sophisticated than simply asking whether an algorithm "clicked accept."

25. Machine Negotiation and Misrepresentation

A machine could theoretically make statements such as:

"The claimant's maximum litigation exposure is AED 700,000."

But if this is generated from unreliable data, the statement could create legal problems.

The system therefore should not be permitted to:

• fabricate evidence;
• falsely state legal rights;
• falsely claim authority;
• manipulate evidence;
• conceal material information where disclosure is legally required.

AI does not create a separate exemption from ordinary civil-law principles.

26. AI Bias in Settlement

Machine-learning systems often operate using historical data.

Suppose historical settlements show that:

Group A settles for 30% of claimed damages.

The AI may use this historical pattern to make lower offers to a new claimant from Group A.

This creates problems of:

• fairness;
• discrimination;
• improper data use;
• statistical bias;
• inadequate individual assessment.

The DIFC's AI Guidance specifically warns about reliability, limitations and potential bias in AI-generated material.

Therefore, settlement systems should not blindly reproduce historical settlement patterns.

27. Human Review as a Governance Safeguard

Even if the legal system permits automated contracting, a sophisticated UAE governance model should distinguish:

Level 1 — Fully automated routine transactions

Small-value, low-risk disputes.

Level 2 — Automated negotiation with human approval

AI negotiates but a human approves the final settlement.

Level 3 — Mandatory human review

Required for:

• high-value claims;
• complex multi-party disputes;
• rights affecting third parties;
• vulnerable parties;
• public-order issues;
• personal-status matters;
• regulatory disputes;
• unusual settlements.

Level 4 — Human-only decision

Certain legally sensitive matters should not be delegated to autonomous systems.

28. Settlement Authority Matrix

A useful UAE corporate model would be:

This is not a statutory UAE table; it is a governance model illustrating how companies could limit delegated authority.

29. Can AI Become a Legal Person?

The concept of machine legal subjectivity remains theoretically controversial.

Three possibilities exist:

Model A — AI as tool

The AI has no independent legal personality.

Model B — AI as authorised agent

The AI acts within authority delegated by a human or company.

Model C — AI as autonomous legal subject

The AI itself possesses legal rights, duties and liability.

For present UAE civil-law analysis, Model B is much more workable than treating the machine as an independent civil person.

The principal remains legally responsible for the authorised activity.

30. Relationship With the DIFC Digital Economy Court

The existence of the DIFC Digital Economy Court is significant because it specifically handles technology-intensive civil and commercial disputes, including AI and blockchain disputes.

This does not, however, mean that the Court automatically recognises AI as an independent legal person.

Instead, it provides a specialised judicial environment for disputes in which technology is central.

31. Important Distinction: AI Settlement vs AI Adjudication

These concepts should never be confused.

AI settlement

The parties voluntarily compromise their dispute.

AI adjudication

A system determines who legally wins.

Settlement is fundamentally based upon consent.

Adjudication is based upon legal authority.

Therefore:

The legal threshold for an AI-assisted settlement is not necessarily the same as the threshold for an AI system making a binding judicial determination.

32. Key Legal Risks

33. Recommended UAE Compliance Architecture

A company using machine-negotiated settlements should ideally establish:

1. Written authority

Specify exactly what the AI can settle.

2. Monetary limits

Set maximum settlement authority.

3. Subject-matter restrictions

Identify disputes that cannot be settled automatically.

4. Approved data sources

Prevent unreliable information from entering the model.

5. Identity verification

Authenticate both parties.

6. Immutable audit trail

Preserve every offer and counteroffer.

7. Model/version logging

Record which AI model produced each decision.

8. Human escalation

Automatically escalate unusual or high-value settlements.

9. Legal review

Require legal approval for significant releases or waivers.

10. Post-settlement monitoring

Check whether the AI remained within its authority.

34. Practical Example

Facts

Company A claims AED 800,000.

Its board authorises an AI settlement system to negotiate claims between AED 300,000 and AED 600,000.

The AI analyses the dispute and negotiates:

• Offer 1: AED 350,000
• Counteroffer: AED 550,000
• AI counteroffer: AED 500,000
• Final acceptance: AED 525,000

The AI automatically generates a settlement agreement.

Legal analysis

Authority:
AED 525,000 falls within the authorised range.

Electronic formation:
Electronic contracting is recognised under Federal Decree-Law No. 46 of 2021, including automated electronic transactions.

Settlement:
The settlement must satisfy the substantive requirements applicable to settlement under the Civil Transactions Law.

Evidence:
The company should preserve the complete negotiation record.

Attribution:
The AI's actions must be attributable to the authorised company/system.

Human input:
A human need not necessarily manually click the final acceptance if the system was validly authorised to conclude such automated transactions; Article 11 is important here.

35. When Human Approval Becomes Especially Important

Human approval should be strongly considered where the settlement:

• releases substantial claims;
• affects third-party rights;
• includes admissions;
• settles multiple jurisdictions;
• involves complex guarantees;
• concerns confidential information;
• affects regulatory rights;
• involves vulnerable persons;
• contains unusual contractual waivers;
• exceeds programmed authority;
• depends on disputed expert evidence.

The DIFC's AI guidance reinforces the broader principle that AI should not replace necessary human decision-making and that AI-generated material requires verification.

36. Mainland UAE vs DIFC

The DIFC is a distinct common-law-influenced jurisdiction and its decisions should not automatically be treated as binding precedents for mainland UAE courts.

37. Six Core Case Principles

38. Key Legal Formula

For examination purposes:

Validity of Machine-Negotiated Settlement = Legal Capacity + Authority + Consent/Attribution + Permissible Subject Matter + Electronic Formation + Evidence + Compliance + Auditability

And:

AI Automation ≠ AI Legal Personality

Algorithmic Acceptance ≠ Automatic Settlement Validity

Electronic Contract ≠ Unlimited AI Authority

Machine Negotiation ≠ Machine Adjudication

39. Conclusion

UAE law provides a significant foundation for automated contracting because Federal Decree-Law No. 46 of 2021 expressly recognises contracts formed between programmed automated electronic systems.

At the same time, a settlement is more legally sensitive than an ordinary automated transaction. The new Civil Transactions Law regulates settlement through specific requirements concerning capacity, subject matter, consideration and proof.

Accordingly, the strongest legal model is not to treat AI as an independent legal person, but as an authorised automated mechanism through which a natural or juridical person acts.

The central principle is:

A machine may negotiate and potentially conclude a settlement without a human manually approving each individual offer, where the system has been properly authorised and the transaction satisfies UAE requirements for automated contracting and settlement. But the absence of human input does not remove the requirements of legal authority, capacity, valid subject matter, attribution, evidence, good faith, confidentiality, data protection and procedural fairness.

The DIFC's Digital Economy Court and its AI guidance demonstrate the UAE's broader movement toward technologically enabled dispute resolution, while simultaneously retaining verification, transparency and human legal responsibility as important safeguards.

Exam conclusion:
Machine-negotiated settlement in UAE law is best understood as authorised automated contracting rather than independent machine legal action. The critical legal issue is not whether a human physically pressed the final button, but whether the automated system was legally authorised, properly attributed, operated within its mandate, and produced a settlement satisfying the substantive and evidentiary requirements of UAE law.