Civil Law and Artificial Intelligence Contract Interpretation Disputes in Europe

1. Introduction

Artificial Intelligence (“AI”) is increasingly involved in the formation, performance, monitoring, modification, and interpretation of contracts. Examples include:

AI-generated contractual clauses;

automated pricing and payment systems;

algorithmic calculation of contractual charges;

AI systems interpreting contractual obligations;

automated renewal or termination;

algorithmic credit and insurance decisions;

smart contracts and code-based contractual performance;

AI-generated invoices;

AI-assisted contract review;

automated interpretation of ambiguous contractual language; and

disputes over whether an AI system acted consistently with the parties' agreement.

A central legal question is:

If an AI system produces an interpretation or contractual outcome that conflicts with the written contract, which prevails—the contractual text, the parties' objectively ascertainable intention, or the AI-generated result?

European civil law generally treats AI as a technological mechanism rather than an independent contracting party. The contractual rights and obligations remain attributable to the human or legal persons who entered into the agreement.

Importantly, there is still limited European case law directly concerning AI interpreting contractual language. The developing legal framework therefore combines traditional contract-interpretation principles with newer CJEU case law on automated decision-making, transparency, consumer contracts and algorithmic processing. The recent Bulgarian Yettel reference is particularly significant because it directly raises questions about AI-generated invoices and contractual obligations, although it was still pending rather than a final judgment at the time of the source material. (EUR-Lex)

2. Meaning of AI Contract Interpretation

AI contract interpretation occurs where an automated or AI system is used to determine the meaning, application, or consequences of contractual provisions.

For example, a contract may state:

“The supplier shall deliver within a reasonable period.”

An AI system may interpret “reasonable period” as 10 days based on historical data, while the parties argue that industry practice requires 20 days.

A dispute can therefore arise concerning:

What did the parties actually agree?

What does the written contract objectively mean?

Can the AI's interpretation become contractually binding?

Was the AI authorised to modify or interpret the contract?

Was the AI's output communicated to the other party?

Was the automated decision transparent?

Can an AI-generated term constitute an unfair consumer term?

Who bears responsibility for an erroneous AI interpretation?

3. Fundamental Civil-Law Principle: AI Does Not Automatically Become a Contracting Party

The starting point remains traditional contract law.

An AI system normally has no independent legal personality merely because it can generate text or make decisions.

Therefore:

AI output ≠ automatically contractual obligation.

The relevant legal person is normally:

the company operating the AI;

the contracting party;

the software provider, where appropriate;

the person who authorised the AI;

or another legally responsible actor depending on the circumstances.

Thus, if a company uses AI to interpret a supply agreement and the AI incorrectly concludes that a delivery deadline has expired, the legal issue is generally whether the company's contractual conduct and the contract itself support that conclusion—not whether the AI independently created a new legal rule.

4. Traditional European Rules of Contract Interpretation

AI does not eliminate conventional principles of interpretation.

Courts generally examine:

A. Contractual wording

The first question is what the actual contractual language says.

B. Common intention

Where relevant, courts consider the parties' common intention at the time of contracting.

C. Objective meaning

Where subjective intention cannot be established, the contract may be interpreted according to the meaning reasonably attributable to the language and circumstances.

D. Context

Courts may examine:

negotiations;

previous dealings;

industry practice;

commercial circumstances;

subsequent conduct;

technical documentation.

E. Good faith

Good faith can restrict opportunistic reliance on an automated interpretation that contradicts the commercial purpose of the agreement.

F. Mandatory consumer protection

A contractual AI system cannot simply override mandatory consumer-protection rules.

This is especially important under Directive 93/13/EEC concerning unfair terms.

5. AI-Generated Interpretation Is Evidence, Not Automatically the Law

An important distinction is:

AI interpretation
↓
may constitute evidence or a contractual performance mechanism

but

AI interpretation
≠
automatically authoritative legal interpretation.

A court can examine:

the original contract;

the AI's instructions;

the data supplied to the system;

the algorithm/model;

the resulting output;

the parties' communications;

industry practice;

applicable legislation.

Therefore, a company normally cannot simply argue:

“Our AI interpreted the clause this way, therefore this is legally what the clause means.”

The court remains responsible for applying the governing law.

6. Contractual AI and the Principle of Transparency

Transparency becomes particularly important where AI determines contractual consequences.

The CJEU's developing automated-decision case law indicates that individuals may have rights to understand the logic behind significant automated decisions.

This is particularly relevant where AI affects:

credit;

insurance;

telecommunications;

employment contracts;

pricing;

access to services;

termination;

contractual penalties.

7. Case Law

Case 1 — SCHUFA, Case C-634/21

OQ v Land Hessen and SCHUFA Holding AG

Court: Court of Justice of the European Union
Case: C-634/21
Date: 7 December 2023

This is one of the most important European decisions concerning automated systems and contractual relationships.

SCHUFA generated a probability value concerning an individual's ability to repay a loan. Banks used that value when deciding whether to grant credit.

The CJEU held that where the probability score plays a determining role in establishing, implementing or terminating a contractual relationship, the creation of that score can itself constitute automated individual decision-making under Article 22 GDPR. (EUR-Lex)

Importance for AI contract disputes

The case demonstrates that an AI/algorithmic system cannot necessarily be treated as a neutral technical tool where its output substantially determines the contractual relationship.

Principle

Where an automated output materially determines whether a contractual relationship is created, performed or terminated, EU automated-decision safeguards may become applicable.

Case 2 — CK v Dun & Bradstreet Austria, Case C-203/22

Court: CJEU
Case: C-203/22
Judgment: 27 February 2025

This is especially important for AI transparency.

CK was refused the conclusion or extension of a mobile-phone contract following an automated credit assessment by Dun & Bradstreet.

The CJEU considered the meaning of the GDPR right to obtain “meaningful information about the logic involved” in automated decision-making.

The Court held that the explanation must identify the procedure and principles actually applied to the person's data in producing the result. A mere complicated mathematical formula or disclosure of a complex algorithm is not necessarily sufficient. (EUR-Lex)

The Court also addressed situations involving trade secrets and third-party personal data, requiring the relevant interests to be balanced by the competent authority or court. (FRA)

Contractual significance

Suppose an AI system determines that:

a consumer has breached a contractual condition;

a payment is due;

a customer should be denied renewal; or

a customer is too risky to continue receiving services.

The affected party may have grounds to challenge the opacity of the automated process where GDPR provisions apply.

Principle

Algorithmic complexity does not by itself eliminate the obligation to provide legally meaningful information.

Case 3 — Kásler and Káslerné Rábai v OTP Jelzálogbank, C-26/13

Court: CJEU
Date: 30 April 2014
Case: C-26/13

The dispute concerned foreign-currency consumer loans and contractual exchange-rate provisions.

The CJEU considered the requirement that contractual terms be drafted in plain and intelligible language under the Unfair Terms Directive. (Infocuria)

The Court's approach is highly relevant to AI-generated contractual clauses.

AI application

Imagine that an AI drafts:

“The exchange rate shall be determined according to the applicable algorithm.”

That phrase may be grammatically understandable but still leave the consumer unable to understand:

what data are used;

how the rate is calculated;

when it changes;

who controls the algorithm;

what economic consequences follow.

The Kásler approach therefore supports examining substantive intelligibility, not merely grammatical readability.

Principle

A contractual clause may need to be sufficiently understandable in its economic and practical operation, not merely technically readable.

Case 4 — Aziz v Caixa d'Estalvis de Catalunya, C-415/11

Court: CJEU
Date: 14 March 2013
Case: C-415/11

The case concerned a mortgage contract and potentially unfair contractual provisions.

The CJEU emphasised the importance of effective judicial review of contractual terms and the protection of consumers against significant contractual imbalance. (Infocuria)

AI relevance

Consider an AI system that automatically applies a contractual termination penalty.

The business may argue:

“The algorithm merely applied the contract.”

The court may nevertheless have to determine:

what the term actually means;

whether it creates a significant imbalance;

whether the consumer could understand it;

whether mandatory consumer protection applies.

Principle

Automation does not prevent a court from examining the substantive fairness and legal effect of contractual provisions.

Case 5 — Banco Español de Crédito, C-618/10

Court: CJEU
Date: 14 June 2012
Case: C-618/10

This case concerned an unfair term relating to late-payment interest.

The CJEU held that a national court could not simply rewrite an unfair contractual term to make it acceptable; the unfair term generally had to be set aside rather than judicially rewritten. (Infocuria)

AI significance

This principle becomes important where an AI system:

generates a penalty clause;

calculates late-payment interest;

modifies a contractual formula;

automatically adjusts a charge.

A court should not assume that it can simply “repair” an AI-generated unfair clause by inventing a different formula.

Principle

An automated contractual mechanism cannot be rescued merely by judicially rewriting an unfair term where EU consumer law requires the term to be disregarded.

Case 6 — Orange România SA v ANSPDCP, C-61/19

Court: CJEU
Date: 11 November 2020
Case: C-61/19

Orange România concerned telecommunications contracts and consent to the processing of identity-document information.

The CJEU held that pre-ticked consent mechanisms did not demonstrate valid consent merely because the customer signed the contract. The Court also considered circumstances affecting the customer's freedom of choice. (curia)

AI contractual significance

AI systems increasingly create:

personalised terms;

consent screens;

automated acceptance processes;

digital contracting interfaces.

The mere fact that an AI system records a customer as having “accepted” something does not necessarily establish valid legal consent.

Principle

Automated recording of contractual acceptance does not eliminate the need to establish genuine and legally valid consent.

8. Case 7 — Verein für Konsumenteninformation v Amazon EU, C-191/15

Court: CJEU
Date: 28 July 2016
Case: C-191/15

This case concerned online consumer contracts and a choice-of-law clause contained in Amazon's standard contractual terms.

The CJEU addressed the interaction between:

consumer protection;

standard contractual terms;

choice of law;

unfair terms;

cross-border online contracting.

(Infocuria)

AI significance

AI-driven platforms increasingly generate or personalise contractual conditions across European markets.

A business cannot avoid mandatory consumer protection simply by allowing an algorithm to insert a foreign-law clause into its terms.

Principle

AI-generated or dynamically generated contractual terms remain subject to applicable conflict-of-laws and mandatory consumer-protection rules.

9. Case 8 — Lintner v UniCredit Bank Hungary, C-511/17

Court: CJEU
Case: C-511/17

The CJEU addressed the responsibility of national courts to examine potentially unfair contractual provisions in consumer agreements.

The case is particularly relevant because the Court's consumer-contract jurisprudence requires courts to consider contractual provisions beyond the precise provision initially challenged where EU law requires such examination. (curia)

AI relevance

An AI system might identify one disputed clause while ignoring another related provision.

A court is not necessarily confined to the AI's categorisation.

Principle

The scope of judicial contractual review is determined by law, not by the limits of an AI contract-review system.

10. The Emerging Yettel Bulgaria Case — C-806/24

This is particularly important because it concerns AI-generated contractual invoices directly.

The Sofia District Court referred questions to the CJEU in proceedings involving Yettel Bulgaria and a consumer.

The dispute concerns invoices automatically generated by a system used under a telecommunications contract.

The questions include whether:

consumers have a right to know how automated contractual invoices are generated;

AI Act Article 86 applies to consumer contracts;

courts may require disclosure of black-box data, source code and algorithms;

automated contractual decisions can receive human judicial review;

AI-generated invoices must be expressed in clear and intelligible language;

automated calculations can determine amounts owed under the contract. (EUR-Lex)

Legal importance

This reference could become particularly significant for the future of European AI-contract litigation because it brings together:

AI Act + Consumer Contract Law + Unfair Terms + Transparency + Judicial Review.

However, it should be treated as a pending reference rather than a final CJEU judgment.

11. AI and the Interpretation of Ambiguous Contractual Terms

Suppose a contract contains:

“Delivery shall take place promptly.”

An AI system interprets “promptly” as five working days.

The supplier says:

“Promptly means within the normal industry delivery period.”

The customer says:

“Promptly means immediately.”

The court would ordinarily examine:

wording;

contractual context;

negotiations;

commercial purpose;

previous dealings;

industry practice;

conduct after contracting;

applicable statutory rules.

The AI's interpretation may be evidence, but it does not automatically replace judicial interpretation.

12. AI and Entire-Agreement Clauses

An important problem occurs where the AI was trained on:

negotiations;

emails;

previous contracts;

internal documents;

industry standards.

Suppose the final agreement says:

“This agreement constitutes the entire agreement between the parties.”

The AI might nevertheless discover an earlier email suggesting a different intention.

The court must determine whether that earlier material can legally be used to interpret the contract.

Therefore:

AI's access to historical information does not automatically make that information part of the contract.

The applicable national rules concerning interpretation and evidence remain important.

13. AI Contract Interpretation and Good Faith

Good faith is particularly important in continental European civil-law systems.

An AI system should not normally be used strategically to produce an interpretation that allows one party to:

escape an agreed obligation;

impose an unexpected penalty;

manipulate a price;

terminate a contract automatically;

exploit an ambiguity;

conceal a contractual change.

For example:

Party A knows that its AI pricing system contains a technical error but continues invoicing customers according to the erroneous interpretation.

The dispute may therefore involve not merely contract interpretation but also:

good faith;

abuse of rights;

unjust enrichment;

contractual cooperation;

duty of information;

culpa in contrahendo or contractual fault, depending on the legal system.

14. AI-Generated Contractual Clauses

AI may itself draft contractual provisions.

This creates several problems.

1. Ambiguity

AI may generate vague language.

2. Inconsistency

Two provisions may contradict each other.

3. Hidden assumptions

The AI may incorporate assumptions that were never agreed by the parties.

4. Hallucinated provisions

An AI drafting system may generate a clause that the parties never intended to include.

5. Translation problems

AI-generated multilingual contracts may produce different meanings in different languages.

6. Regulatory incompatibility

An AI may produce a provision inconsistent with mandatory European legislation.

15. Human Intention Versus Machine Output

A fundamental issue can be expressed as:

Party intention → Contract → AI interpretation

not:

AI interpretation → Contractual intention

Unless the parties have expressly incorporated an automated mechanism into their agreement, the AI's later interpretation normally cannot retroactively redefine what was agreed.

This is particularly important where the AI was introduced after the contract was signed.

16. AI and Smart Contracts

Smart contracts create a related but distinct problem.

A smart contract may automatically execute:

IF condition X occurs → payment Y is made.

The code can therefore perform the contract automatically.

But legal interpretation may still be required when:

the code contains a bug;

the code contradicts the natural-language agreement;

an external data feed is incorrect;

fraud occurred;

force majeure applies;

performance becomes impossible;

the parties mutually agreed to amend the arrangement.

European legal analysis generally treats smart contracts as contractual arrangements subject to the applicable legal system rather than as arrangements completely outside ordinary contract law. (EBRD)

17. Code Versus Natural-Language Contract

A particularly difficult dispute is:

Natural-language contract

“Payment shall be made when the shipment arrives.”

Code

IF GPS signal = delivery location THEN release payment

The GPS signal incorrectly identifies the location.

The code releases the money.

The parties dispute whether payment was legally due.

The court may have to determine:

whether the code was incorporated into the contract;

whether the code represents the parties' common intention;

whether the code contains an error;

whether the natural-language text prevails;

whether contractual good faith requires correction;

whether restitution is available.

18. AI and Consumer Contracts

This is probably the most important area of future litigation.

AI can automatically:

generate prices;

calculate penalties;

determine eligibility;

recommend contract renewal;

determine creditworthiness;

create invoices;

modify offers;

determine cancellation charges.

Consumer law imposes important restrictions.

The Kásler jurisprudence establishes the importance of intelligibility. (Infocuria)

The Banco Español de Crédito jurisprudence establishes limits on judicial modification of unfair terms. (Infocuria)

The Aziz line of authority reinforces effective judicial protection against unfair contractual provisions. (Infocuria)

Together, these principles create a significant constraint on opaque AI-based contractual mechanisms.

19. AI and Evidence in Contract Litigation

A party challenging an AI interpretation may seek:

source code;

system documentation;

model version;

prompts;

input data;

output logs;

audit trails;

training information;

configuration files;

decision rules;

human overrides;

version history.

But disclosure is not necessarily unlimited.

CK v Dun & Bradstreet demonstrates the need to balance explanation rights with competing interests such as trade secrets and third-party data. (EUR-Lex)

Thus:

AI transparency does not necessarily mean unrestricted disclosure of source code.

The legally relevant question is often whether sufficient information can be provided to permit effective understanding and judicial review.

20. Burden of Proof

In an AI contract dispute, the claimant may need to establish:

existence of the contract;

relevant contractual provision;

AI-generated interpretation or decision;

divergence between the AI result and contractual obligation;

resulting loss or legal consequence.

The defendant may then need to explain:

how the automated system operated;

what contractual data it used;

whether human review occurred;

whether the system complied with contractual rules.

The precise allocation of evidential burdens depends on national procedural law and the particular EU legislation involved.

21. AI Errors and Contractual Liability

An erroneous AI interpretation can potentially generate liability through several routes.

A. Breach of contract

If the AI causes the company to perform incorrectly.

B. Negligent performance

If reasonable safeguards were not implemented.

C. Misrepresentation

If AI-generated information induced the other party to contract.

D. Unjust enrichment

If an erroneous automated calculation caused excessive payment.

E. Consumer-law remedies

Where an automated term or decision violates mandatory consumer legislation.

F. Data-protection remedies

Where personal-data processing and automated decision-making violate GDPR requirements.

22. AI Contract Interpretation and GDPR Article 22

Article 22 GDPR is particularly significant where AI makes decisions affecting contractual relationships.

The CJEU in SCHUFA held that an automated probability score could itself constitute automated individual decision-making where a third party relied strongly upon it to establish, implement or terminate a contractual relationship. (EUR-Lex)

Article 22 also provides safeguards concerning:

human intervention;

expressing one's point of view;

contesting the decision;

certain restrictions on special-category data.

(EUR-Lex)

Therefore, a company cannot necessarily avoid GDPR obligations by arguing:

“The final contractual decision was technically made by our employee; the employee merely followed the AI score.”

The factual role of the algorithm may matter.

23. AI Contract Interpretation and the EU AI Act

The EU AI Act adds another layer to the legal environment.

For contractual disputes, relevant questions include:

Was the AI system covered by the AI Act?

What role did the provider and deployer have?

Was the system used for a regulated high-risk purpose?

Were transparency obligations triggered?

Was the affected person informed?

Can the system's output be challenged?

How does the AI Act interact with GDPR and consumer law?

The pending Yettel Bulgaria reference illustrates how these regimes may increasingly intersect with ordinary consumer-contract disputes. (EUR-Lex)

24. Cross-Border European Contracts

AI contract disputes frequently involve several countries.

For example:

French customer + German company + Irish AI provider + Dutch cloud infrastructure.

Questions may include:

Which law governs the contract?

Which court has jurisdiction?

Where did the automated decision occur?

Where was the loss suffered?

Which consumer protections apply?

Which country's mandatory rules apply?

The Amazon EU case illustrates the importance of EU conflict-of-laws and consumer-protection rules in cross-border online contracts. (Infocuria)

25. Main Legal Tests for an AI Contract Interpretation Dispute

A European court can conceptually work through the following sequence:

Test 1 — Identify the contract

What is the actual legally binding agreement?

Test 2 — Identify the relevant clause

What provision is being interpreted?

Test 3 — Determine the agreed meaning

What meaning follows from applicable contract law?

Test 4 — Identify the AI's role

Was AI:

drafting,

recommending,

interpreting,

calculating,

executing,

or making the contractual decision?

Test 5 — Determine incorporation

Did the parties agree that AI/code would determine the contractual outcome?

Test 6 — Examine transparency

Could the affected party reasonably understand the automated mechanism?

Test 7 — Apply mandatory law

Check:

consumer law;

GDPR;

AI Act;

unfair-terms legislation;

competition law where relevant;

sector-specific regulation.

Test 8 — Examine evidence

What did the AI actually do?

Test 9 — Determine responsibility

Who controlled or deployed the system?

Test 10 — Determine remedy

Possible remedies include:

damages;

restitution;

specific performance;

declaration of contractual meaning;

setting aside an unfair term;

correction of an automated decision;

termination;

injunction.

26. Key Distinction: AI-Assisted vs AI-Determined Contracts

27. Comparative European Position

France

Contractual interpretation remains rooted in the Civil Code, especially principles concerning contractual force, good faith and interpretation of agreements. AI does not replace judicial interpretation.

Germany

The BGB framework places importance on contractual interpretation, good faith and consumer protection. AI-generated terms remain subject to ordinary contractual requirements.

Spain

Spanish contract law and EU consumer law are particularly relevant to automated consumer-contract disputes, with Aziz and Banco Español de Crédito originating from Spanish proceedings.

Hungary

Kásler demonstrates the importance of intelligibility and consumer understanding in contractual interpretation.

Austria

CK v Dun & Bradstreet Austria demonstrates the increasing importance of algorithmic transparency where automated assessments affect contractual relationships.

Bulgaria

The pending Yettel Bulgaria reference is directly concerned with automated contractual invoices and the interaction between AI, consumer contracts and judicial review. (EUR-Lex)

28. Six+ Key Cases — Quick Revision Table

The first eight are decided cases; Yettel Bulgaria is a pending preliminary reference, so it should not be cited as an established final holding. (EUR-Lex)

29. Major Legal Issues for Future European Litigation

Future cases are likely to concentrate on:

AI-generated contractual clauses

AI interpretation of ambiguous language

Algorithmic price calculation

Automated contractual penalties

AI-generated invoices

Automatic termination

AI-based contract renewal

Smart-contract coding errors

Natural-language/code conflicts

Algorithmic consumer discrimination

AI transparency

Trade-secret protection

GDPR Article 22

EU AI Act Article 86

Human review

Evidence and source-code disclosure

Cross-border jurisdiction

Good-faith obligations

Liability for AI errors

Remedies for automated contractual decisions

30. Conclusion

European civil law does not generally allow an AI system to become the final source of contractual meaning merely because the parties use it.

The traditional hierarchy remains important:

Contract → Applicable law → Parties' legally relevant intention/context → Judicial interpretation → AI output as evidence or contractual mechanism where properly incorporated.

The emerging CJEU jurisprudence adds a second layer:

AI/algorithmic decision → Transparency → Explainability → Human/legal review → Consumer protection → GDPR/AI Act compliance.

The most important current authorities are SCHUFA on automated decisions affecting contractual relationships, CK v Dun & Bradstreet on meaningful explanation of algorithmic logic, and Kásler, Aziz, Banco Español de Crédito, Orange România, Amazon EU and Lintner on contractual intelligibility, fairness, consent and judicial protection. (EUR-Lex)

Exam Formula

AI Contract Interpretation Dispute =

Contract Text + Parties' Intention + Good Faith + Context + AI Role + Transparency + Consumer Protection + GDPR + AI Act + Evidence + Human Review + Remedy.