Civil Law and Airline AI Pricing Algorithm Discrimination Claims in Europe

1. Introduction

Airline AI pricing algorithm discrimination claims arise when an airline, online travel intermediary, loyalty platform, or revenue-management system uses automated models to determine or personalise ticket prices, ancillary charges, upgrades, discounts, or offers, and the resulting treatment allegedly disadvantages passengers because of a protected characteristic or through a discriminatory proxy.

A typical system might analyse:

search history;

booking history;

location;

device;

browsing behaviour;

loyalty status;

purchasing patterns;

timing of purchase;

route;

language;

customer profile;

inferred willingness to pay;

previous expenditure; and

other behavioural data.

The important legal point is that different prices are not automatically unlawful discrimination. Airline pricing in the EU is generally commercially flexible, but that freedom operates alongside consumer-protection, GDPR, anti-discrimination and, in appropriate circumstances, competition-law requirements. The CJEU has specifically recognised airline pricing freedom under Regulation 1008/2008 while also confirming that consumer-protection rules can apply.

There is currently no major CJEU judgment directly deciding an AI airline-ticket pricing discrimination case. Therefore, the cases below are a combination of direct airline-pricing authorities and highly relevant AI, profiling, GDPR and discrimination authorities. They should not be presented as if the CJEU has already decided that a particular airline AI pricing practice is discriminatory.

2. Basic Legal Structure

An airline AI pricing dispute can involve several legal regimes simultaneously:

A. EU aviation/pricing law

Especially Regulation (EC) No 1008/2008.

B. GDPR

Particularly:

profiling;

automated decision-making;

lawful processing;

transparency;

fairness;

data minimisation;

access rights.

C. EU anti-discrimination law

Depending upon the protected characteristic and context:

race/ethnic origin;

sex;

nationality in appropriate contexts;

disability;

age;

religion/belief;

sexual orientation.

D. Consumer law

Particularly:

unfair commercial practices;

unfair contractual terms;

price transparency;

misleading information.

E. Competition law

Potentially:

Article 102 TFEU;

discriminatory pricing by a dominant undertaking;

exploitation of personal data;

platform-related conduct.

F. National civil law

Depending on the Member State:

tort/delict;

contractual liability;

restitution;

damages;

injunctions;

non-material damages.

3. What Is AI Pricing Discrimination?

Consider two passengers:

Passenger A

Ticket price = €180.

Passenger B

Ticket price = €260.

A difference alone does not establish discrimination.

The claimant must investigate:

Why did the algorithm produce different prices?

For example, if the model uses:

postal code → inferred ethnicity → higher price

or:

disability-related travel information → higher price

or:

gender-related behavioural proxy → systematically higher price,

a discrimination claim becomes substantially more serious.

4. Direct and Indirect Algorithmic Discrimination

Direct discrimination

The algorithm expressly or effectively uses a protected characteristic.

Example:

“Passenger's sex = female → increase price by 10%.”

That is relatively straightforward.

Indirect discrimination

The algorithm does not expressly use the protected characteristic but uses a proxy.

Example:

geographic area → socioeconomic characteristics → ethnic composition → higher price.

The airline may argue:

“The algorithm only uses location.”

The claimant may respond:

“Location functions as a proxy for ethnic origin.”

This is where the CJEU's indirect-discrimination jurisprudence becomes important.

5. Case 1 — CHEZ Razpredelenie Bulgaria, C-83/14

CJEU, 16 July 2015

This is one of the most important cases for understanding proxy discrimination.

An electricity company installed electricity meters at unusually high locations in neighbourhoods predominantly inhabited by Roma people. The company argued that the measure was intended to prevent fraud.

The CJEU examined both direct and indirect discrimination under Directive 2000/43. It explained that apparently neutral criteria may constitute indirect discrimination where they place a particular ethnic group at a particular disadvantage, subject to objective justification and proportionality. (Infocuria)

Application to airline AI

Suppose an airline algorithm uses:

postcode + purchasing behaviour + language

rather than ethnicity itself.

If those variables systematically disadvantage a particular ethnic group, CHEZ provides a useful framework for analysing proxy/indirect discrimination.

Important principle

An algorithm does not escape discrimination law merely because it does not explicitly contain the protected characteristic.

However, the claimant still has to establish the legally required disadvantage and, where relevant, the conditions for indirect discrimination.

6. Case 2 — Feryn, C-54/07

CJEU, 10 July 2008

In Feryn, a company publicly indicated that it would not recruit workers of a particular ethnic origin.

The CJEU held that public statements indicating discriminatory recruitment policy can constitute direct discrimination even without an identifiable individual applicant. (curia)

Relevance to AI pricing

The context was employment rather than airline pricing, so this is analogical rather than direct authority.

Its significance is evidential.

An airline may say:

“Nobody instructed the algorithm to discriminate.”

But evidence concerning:

algorithm design;

business instructions;

pricing policies;

model objectives;

data selection;

internal documentation;

may help determine whether discriminatory treatment was intentionally embedded in the system.

Key principle

Discriminatory conduct can be established through surrounding evidence rather than a literal written instruction saying “discriminate.”

7. Case 3 — Asociația ACCEPT, C-81/12

CJEU, 25 April 2013

The case concerned discriminatory statements relating to recruitment based on sexual orientation.

The CJEU addressed the evidentiary framework under which facts creating an appearance of discrimination can shift the burden of proof, subject to the applicable national and EU rules. (Infocuria)

Application to AI pricing

This principle can become important where a passenger cannot access the airline's algorithm.

Suppose a claimant demonstrates:

a large statistical price difference;

a protected characteristic;

repeated treatment across transactions;

similar booking conditions;

no obvious legitimate pricing explanation.

The question becomes:

What evidence is sufficient to create an inference of discrimination?

Algorithmic discrimination litigation therefore makes evidence and burden of proof especially important.

8. Case 4 — SCHUFA Holding, C-634/21

CJEU, 7 December 2023

This is a major automated-decision authority.

SCHUFA generated credit scores automatically. Those scores were then used in decisions concerning individuals.

The CJEU interpreted Article 22 GDPR concerning automated individual decision-making and profiling. (Infocuria)

Why this matters to airline pricing

Imagine:

Airline AI → individual passenger score → personalised ticket price.

If the price determination is sufficiently consequential and involves automated processing of personal data, GDPR Article 22 and related provisions may become relevant.

The precise application depends on the facts, including whether there is an “automated decision” producing legal effects or similarly significant effects.

Important distinction

Not every automated price calculation automatically falls under Article 22.

A court would need to examine:

how the decision is made;

whether human intervention is meaningful;

whether the price has sufficiently significant effects;

what personal data are processed;

the applicable legal basis.

Principle

Automation does not place consequential decisions outside GDPR scrutiny.

9. Case 5 — Dun & Bradstreet Austria, C-203/22

CJEU, 27 February 2025

This is especially important for AI pricing disputes because it concerns algorithmic explanation.

Dun & Bradstreet generated an automated credit assessment.

The CJEU held that the data subject is entitled to information about the logic involved in automated decision-making that is sufficiently meaningful to enable the person to understand and challenge the decision. (curia)

Application to airline pricing

Suppose an airline tells a passenger:

“Our AI determined that your ticket should cost €420.”

That statement alone may be inadequate where GDPR rights to meaningful information are applicable.

The claimant may need information capable of explaining:

relevant categories of data;

principal factors;

how those factors affected the result;

the significance of the relevant factors;

how the individual result was generated.

Trade secrets

The airline cannot necessarily be required to publish its entire source code.

Dun & Bradstreet recognises the need to balance explanation rights against matters such as trade secrets and the rights of others. (Curia)

Core principle

“Black box” cannot automatically mean “no explanation.”

10. Case 6 — Meta Platforms, C-252/21

CJEU, 4 July 2023

The case involved Meta's combination of personal data from Facebook and other sources.

The CJEU examined the relationship between:

GDPR;

processing of personal data;

competition law;

dominance.

The Court confirmed that a national competition authority may take account of GDPR compliance when examining an abuse of dominance, while respecting the institutional role of data-protection authorities. (Infocuria)

Airline relevance

Suppose a dominant airline or airline platform collects extensive passenger data and uses it for personalised pricing.

The dispute could potentially involve:

Data processing + market power + discriminatory pricing

Thus, GDPR and competition law are not necessarily isolated legal compartments.

Principle

Data-protection compliance can be relevant to competition-law analysis in an appropriate dominance case.

11. Case 7 — Planet49, C-673/17

CJEU, 1 October 2019

Planet49 concerned online consent and cookies.

The CJEU held that consent for storing/accessing information could not simply be obtained through a pre-ticked checkbox. Active consent requirements were important. (curia)

Airline pricing relevance

Airlines may use:

cookies;

tracking pixels;

account information;

browsing history;

device identifiers;

loyalty-program data.

If such data are used to personalise prices or offers, the legality of the underlying data collection and processing may become relevant.

Important qualification

Planet49 does not establish that personalised pricing is discriminatory.

It establishes a principle concerning lawful consent and electronic tracking.

Therefore it is supporting/analogical authority, not a direct airline AI-pricing case.

12. Case 8 — Vueling Airlines, C-487/12

CJEU, 18 September 2014

This is directly relevant to airline pricing.

Vueling concerned the relationship between airline pricing freedom and consumer protection.

The CJEU recognised that Regulation 1008/2008 provides airlines with freedom to set air fares, while also confirming the relevance of consumer-protection rules and transparency requirements. (Infocuria)

Application

An airline may generally differentiate prices based on legitimate commercial factors.

For example:

booking time;

seat availability;

fare class;

flexibility;

route;

demand.

But pricing freedom does not mean:

“Every algorithmically generated price is legally immune from scrutiny.”

Consumer-protection and other mandatory EU rules continue to apply.

13. Regulation 1008/2008 and Airline Pricing

Article 22(1) of Regulation 1008/2008 establishes pricing freedom for air services.

However, Article 23 establishes transparency requirements concerning the final price.

The CJEU has repeatedly treated those provisions as important to ensuring consumers can compare fares.

In Vueling, the Court explained that airline pricing freedom does not exclude consumer-protection rules.

Therefore:

Airline AI pricing

Commercial pricing freedom

price transparency

consumer protection

GDPR

anti-discrimination law

=

legally constrained algorithmic pricing

14. What Counts as Discriminatory Pricing?

A claimant should distinguish at least four situations.

A. Legitimate dynamic pricing

Example:

Ticket increases from €150 to €250 because seats are being sold quickly.

Normally this is not, by itself, discrimination.

B. Personalised pricing

Example:

Returning customer receives a different offer based on loyalty status.

Different treatment is not automatically unlawful.

C. Protected-characteristic discrimination

Example:

Algorithm charges passengers differently because of ethnic origin.

Potential discrimination claim.

D. Proxy discrimination

Example:

Algorithm uses location or behavioural variables that function as a proxy for ethnic origin and systematically disadvantage that group.

This is legally more complicated and brings CHEZ-type indirect discrimination principles into the analysis. (Infocuria)

15. Algorithmic Proxy Variables

Common potential proxies include:

A variable is not automatically unlawful merely because it correlates with a protected characteristic.

The legal question is whether its use produces treatment prohibited by the applicable discrimination regime and whether any available justification satisfies the relevant legal test.

16. Sensitive Personal Data

GDPR Article 9 gives special protection to certain categories of personal data, including information concerning:

racial or ethnic origin;

health;

religious or philosophical beliefs;

sexual orientation;

biometric data for unique identification.

If an airline AI system processes such information, additional GDPR restrictions may apply.

This is particularly important for:

Disability-related pricing

Suppose an airline's algorithm uses information concerning:

wheelchair assistance + medical assistance + health-related travel requirements

to determine the price.

The legal analysis cannot simply be:

“It is only a pricing variable.”

The system must also be analysed under the GDPR's special-category rules where applicable.

17. Automated Decision-Making

A particularly important distinction is:

Automated recommendation

AI recommends:

€350.

Human employee decides:

€350.

versus:

Automated decision

AI directly determines:

Passenger must pay €350.

The second situation potentially raises stronger Article 22 issues.

SCHUFA is therefore highly relevant to determining whether an automated system effectively makes a consequential decision. (Infocuria)

18. Human Intervention

Simply putting a human somewhere in the process does not necessarily make the decision genuinely human.

For example:

AI generates price → employee clicks “approve” automatically.

A court may need to examine whether that employee actually:

reviews the result;

has authority to change it;

considers the passenger's circumstances;

understands the relevant factors;

can reject the algorithm's recommendation.

This is particularly important when GDPR Article 22 is invoked.

19. Evidence in AI Pricing Litigation

Evidence may include:

Technical evidence

model architecture;

model documentation;

feature lists;

training datasets;

validation reports;

audit logs;

model outputs.

Commercial evidence

pricing policies;

revenue-management instructions;

business objectives;

customer-segmentation documents.

Statistical evidence

price distributions;

repeat observations;

control groups;

protected-group comparisons;

regression analysis.

GDPR evidence

privacy notices;

consent records;

processing registers;

DPIAs;

automated-decision explanations.

20. Statistical Evidence

Suppose 10,000 comparable bookings are examined.

This is evidence of a disparity.

But:

Disparity ≠ automatic proof of unlawful discrimination.

The airline may demonstrate that the difference results from:

booking timing;

route;

fare class;

demand;

cancellation flexibility;

baggage;

seat selection;

loyalty benefits;

other legitimate variables.

Therefore the claimant must investigate causation and the legal discrimination test, not merely price differences.

21. Burden of Proof

In applicable EU equality regimes, once sufficient facts establish an appearance or presumption of discrimination, evidential burdens may shift.

Asociația ACCEPT is useful here because the CJEU addressed circumstances capable of generating an inference of discrimination and the resulting burden-of-proof structure. (Infocuria)

For AI litigation, this becomes:

Statistical evidence + algorithmic evidence + protected characteristic + comparable transactions → possible inference → airline explanation/justification

The exact burden depends upon the applicable discrimination directive and national implementing law.

22. Consumer Law Dimension

Even where discrimination cannot be established, AI pricing may raise consumer-law questions.

Potential problems include:

misleading presentation of price;

hidden charges;

deceptive urgency;

unclear personalised pricing;

failure to disclose mandatory charges;

unfair terms;

manipulation of consumer choice.

Vueling demonstrates that airline pricing freedom operates alongside consumer protection.

23. Transparency of Final Price

Airline AI systems might display:

“Your price: €399.”

The system may actually calculate:

Base fare €230

dynamic adjustment €80

personalised adjustment €50

baggage €39

The legal question becomes whether the passenger receives the information required by EU airline pricing rules.

Article 23 of Regulation 1008/2008 is important because the CJEU has emphasised the need to disclose relevant components of the final price. (Infocuria)

24. Competition-Law Dimension

Suppose an airline is dominant on a particular route.

It uses AI to charge:

Passenger Group A = €200
Passenger Group B = €450.

Competition law may become relevant if the conduct involves a prohibited form of discriminatory or exploitative conduct by a dominant undertaking.

But:

Different prices alone do not establish Article 102 abuse.

The relevant market, dominance, conduct, effects and legal test must be established.

Meta Platforms is important for the broader proposition that GDPR-related conduct can form part of competition-law analysis where dominance is being examined. (curia)

25. Contractual Civil Liability

A passenger may potentially bring a civil claim based on:

breach of contract;

unfair contractual terms;

statutory consumer rights;

discrimination legislation;

GDPR;

national tort/delict;

unjust enrichment.

The exact remedy depends on national law.

Possible remedies include:

price difference;

compensation;

restitution;

non-material damages;

injunction;

correction/deletion of data;

cessation of discriminatory processing;

regulatory penalties.

26. GDPR Damages

Where GDPR requirements are violated, Article 82 GDPR provides a framework for compensation for material or non-material damage.

An important issue is therefore:

Did the algorithm merely generate an unwanted price, or did unlawful personal-data processing cause legally compensable damage?

This requires separate analysis.

Not every GDPR infringement automatically establishes a particular amount of civil damages.

27. AI Act Dimension

The EU AI Act, Regulation (EU) 2024/1689, adds another regulatory layer.

However, an important distinction must be maintained:

AI Act violation ≠ automatic civil damages liability.

The AI Act is principally a regulatory framework concerning AI systems according to their risk classification.

For airline pricing, the relevant question is first:

What type of AI system is being used, and does the AI Act classify that particular use as prohibited, high-risk, or subject to another category of obligation?

An ordinary revenue-management system should not automatically be labelled “high-risk AI” merely because it uses AI.

28. AI Act and Discrimination

Where an AI system falls within a category subject to specific obligations, the AI Act's requirements concerning matters such as:

risk management;

data governance;

transparency;

record-keeping;

accuracy;

robustness;

cybersecurity;

human oversight;

may become relevant.

These requirements can also create valuable evidence for civil litigation.

For example:

poor-quality training data → biased model → discriminatory output

could support arguments concerning negligent design or inadequate governance, depending upon the applicable national civil law.

29. Important Distinction: Price Discrimination vs Dynamic Pricing

Dynamic pricing

Price changes because:

demand changes;

inventory changes;

departure date approaches.

Personalised pricing

Price differs because:

user history;

account characteristics;

browsing behaviour;

loyalty profile.

Discriminatory pricing

Price differs because of a legally protected characteristic or an unlawful proxy producing prohibited unequal treatment.

Therefore:

Dynamic ≠ personalised ≠ discriminatory

This distinction is essential in examination answers.

30. Six+ Case-Law Summary

The CJEU sources confirm the holdings of these cases. (Infocuria)

31. Hypothetical Example

Assume an airline operates an AI revenue-management system.

It considers:

route;

departure date;

remaining seats;

passenger history;

postcode;

device;

loyalty status;

browsing behaviour.

A passenger discovers that people from a particular ethnic community systematically pay 15–20% more for otherwise comparable tickets.

Legal analysis

Step 1 — Pricing freedom

Airline pricing freedom exists under Regulation 1008/2008.

↓

Step 2 — Data

Determine what personal data the AI processes.

↓

Step 3 — Profiling

Determine whether passenger profiling occurs.

↓

Step 4 — Automated decision

Assess whether Article 22 GDPR is engaged.

↓

Step 5 — Explanation

Dun & Bradstreet becomes relevant.

↓

Step 6 — Discrimination

Apply direct/indirect discrimination principles.

↓

Step 7 — Proxy

Examine whether postcode/location functions as a discriminatory proxy.

↓

Step 8 — Justification

Assess the airline's legitimate commercial explanation and, where applicable, proportionality.

↓

Step 9 — Consumer protection

Check whether the final price and mandatory components were transparently presented.

↓

Step 10 — Civil remedy

Determine damages, restitution, injunction or other relief under applicable national law.

32. Defences Available to the Airline

An airline might argue:

1. No protected characteristic was used

The algorithm used only commercial variables.

2. No discriminatory effect

Statistical differences disappear after controlling for legitimate pricing factors.

3. Objective justification

The variable is genuinely connected with a legitimate commercial purpose.

4. No automated decision under Article 22

A meaningful human decision-maker intervened.

5. No unlawful processing

The personal data were processed on an appropriate GDPR legal basis.

6. No causal connection

The alleged protected characteristic did not cause the price difference.

7. No compensable damage

Even if a regulatory infringement occurred, the claimant must establish the required damage for a civil compensation claim.

33. Claimant's Evidence Strategy

A claimant would typically seek to establish:

Protected characteristic → algorithmic variable/proxy → statistically significant disparity → comparable passengers → algorithmic causation → lack of lawful justification → financial/non-material damage

Useful evidence may include:

booking screenshots;

price histories;

repeated searches;

identical itinerary comparisons;

data-access requests;

algorithmic explanations;

expert statistical evidence;

privacy notices;

loyalty-program terms;

airline pricing policies.

34. Most Important Legal Problem: Black-Box Evidence

The central difficulty is:

How can a passenger prove algorithmic discrimination without seeing the algorithm?

This is where Dun & Bradstreet becomes particularly significant.

The claimant may need enough information to understand:

which factors were used;

which factors mattered;

how they affected the result;

whether the result can be challenged.

But the airline's trade secrets must also be considered. The CJEU's approach requires balancing explanation rights against protected information. (Curia)

35. Civil Liability Formula

AIRLINE AI PRICING DISCRIMINATION LIABILITY

AI SYSTEM

↓

PERSONAL DATA

↓

PROFILING

↓

PRICE DECISION

↓

PROTECTED CHARACTERISTIC / PROXY

↓

DISPARATE TREATMENT

↓

DISCRIMINATION TEST

↓

JUSTIFICATION / PROPORTIONALITY

↓

GDPR + CONSUMER + AVIATION LAW

↓

CAUSATION

↓

DAMAGE

↓

CIVIL REMEDY

36. Ultra-Basic Exam Formula

Remember:

“PRICE → DATA → AI → PROXY → DISPARITY → JUSTIFICATION → DAMAGE”

PRICE = airline fare

DATA = passenger information

AI = automated pricing/profiling

PROXY = variable correlated with protected characteristic

DISPARITY = different treatment

JUSTIFICATION = legitimate and proportionate reason where legally relevant

DAMAGE = financial/non-material harm

37. Final Conclusion

European airline AI pricing discrimination litigation sits at the intersection of airline pricing freedom, consumer protection, GDPR, equality law, AI regulation and national civil liability.

The most important point is that there is presently no established CJEU rule saying that AI-based personalised airline pricing is itself unlawful. The legality depends upon what data the system uses, whether a protected characteristic or proxy is involved, whether the system makes a legally significant automated decision, whether the pricing difference is objectively justified where required, and whether EU and national transparency, consumer, data-protection and discrimination rules have been complied with.

The strongest case-law combination is:

Vueling → establishes airline pricing freedom and consumer-law limits. (Infocuria)
CHEZ → explains indirect/proxy discrimination. (Infocuria)
SCHUFA → addresses consequential automated decisions. (Infocuria)
Dun & Bradstreet → strengthens meaningful explanation rights. (Infocuria)
Meta Platforms → connects GDPR and competition analysis. (Infocuria)
Asociația ACCEPT/Feryn → provide useful discrimination-evidence and burden-of-proof principles. (Infocuria)

Exam conclusion:

“An airline's AI pricing freedom is not absolute. Where automated pricing uses personal data or discriminatory proxies and produces prohibited unequal treatment, liability may arise through the combined operation of EU aviation pricing rules, GDPR, anti-discrimination law, consumer protection, competition law and national civil liability. The central litigation questions are algorithmic causation, proxy discrimination, justification, transparency, proof and damage.”