Artificial Moral Agency Claims in Europe

1. Meaning of Artificial Moral Agency

Artificial Moral Agency (AMA) refers to the idea that an artificial intelligence system may be capable of making decisions that appear to involve moral judgment, ethical choice, responsibility, or value-sensitive behaviour.

Examples include an AI system that:

decides whom to admit to a university;

determines whether a person receives a loan or insurance;

prioritises patients for medical treatment;

recommends or rejects employment candidates;

controls an autonomous vehicle in a dangerous situation;

determines whether online content should be removed;

allocates welfare or public resources;

makes policing or surveillance recommendations;

autonomously trades financial assets;

determines whether a user should be blocked or suspended.

The important legal question is:

Can an AI system itself become the legal or moral bearer of responsibility for a harmful decision, or does responsibility remain with the human or legal entity that designed, supplied, deployed, controlled, or benefited from it?

As European law currently stands, AI systems are generally not recognised as independent moral or legal agents possessing their own enforceable rights and duties. Responsibility ordinarily remains with natural persons, companies, public authorities, developers, deployers, employers, manufacturers, or other legally recognised actors.

There is therefore no general European cause of action called an “artificial moral agency claim.” Such claims are normally constructed through existing areas of law.

2. Artificial Moral Agency Versus Artificial Legal Personhood

These concepts must be carefully distinguished.

An AI can be autonomous in a technical sense without being a legal person.

For example, an autonomous vehicle may independently select braking or steering actions. That does not automatically mean that the vehicle itself is legally responsible for the resulting accident.

3. Present European Legal Position

The current European approach is essentially based on human and organisational accountability.

The EU's regulatory framework generally treats AI as a technology that is:

developed by providers;

supplied or placed on the market;

deployed by organisations;

operated by humans or organisations;

subject to risk-management and governance obligations.

The EU AI Act does not create a general legal personality for AI systems.

Consequently, an AI cannot presently be treated in Europe in the same manner as:

a natural person;

a corporation;

a foundation;

a public authority.

The central question therefore becomes:

Which human or legal entity exercised the legally relevant control, created the foreseeable risk, failed to comply with a duty, or benefited from the AI system?

4. What Would an Artificial Moral Agency Claim Look Like?

A claimant might argue:

the AI acted autonomously;

the AI made a morally significant decision;

the decision caused legally recognised harm;

the system was sufficiently autonomous that ordinary human control was limited;

the developer/deployer should nevertheless be accountable for creating or operating the system;

existing law should attribute responsibility to a human or corporate actor.

A court would ordinarily not stop at the AI itself.

It would investigate:

AI system → developer → provider → deployer → operator → decision → harm → legal duty → causation → remedy.

5. Principal Categories of Artificial Moral Agency Claims

A. AI Decision-Making Claims

AI may make or substantially influence decisions concerning:

employment;

education;

credit;

insurance;

healthcare;

welfare;

immigration;

policing;

taxation;

housing.

A claimant may challenge the decision under:

GDPR;

equality law;

administrative law;

consumer law;

contract law;

negligence/delict;

fundamental rights.

B. Autonomous Vehicle Claims

An autonomous vehicle may independently determine:

braking;

acceleration;

lane changes;

collision avoidance;

route selection.

The fact that the vehicle makes the immediate decision does not normally make it a legal person.

Potential defendants include:

manufacturer;

software provider;

component manufacturer;

owner;

operator;

maintenance provider.

Product liability and negligence principles therefore remain central.

C. Medical AI Claims

Medical AI can recommend:

diagnosis;

treatment;

surgery;

medication;

triage;

patient prioritisation.

If an AI recommendation causes injury, possible claims may involve:

defective product;

medical negligence;

professional negligence;

failure to warn;

inadequate human supervision;

data protection;

discrimination.

The physician or healthcare institution cannot automatically avoid responsibility merely by saying that the algorithm generated the recommendation.

6. AI and Moral Responsibility

A philosophical distinction is important.

Moral agency

A moral agent traditionally possesses capacities such as:

understanding;

intention;

choice;

awareness;

reasons for action;

ability to recognise obligations;

ability to be praised or blamed.

Legal agency

Legal agency concerns whether the law recognises an entity as capable of:

holding rights;

undertaking obligations;

entering legal relationships;

being sued;

being liable.

An AI may exhibit behaviour resembling moral reasoning without possessing legal personality.

This is one of the most important distinctions in European AI liability.

7. Case Law

There is no established European case-law doctrine granting AI systems independent moral agency. The following cases are therefore best divided into directly relevant AI/data cases and analogical human-rights, privacy, discrimination and technology cases.

Case 1: SCHUFA – Joined Cases C-26/22 and C-64/22

SCHUFA Holding (Scoring), C-26/22 and C-64/22, CJEU, 7 December 2023

Principle

The CJEU examined automated scoring under GDPR Article 22.

The case is extremely important for artificial moral agency because it demonstrates that an algorithmically generated score can have legally significant consequences even though the ultimate decision may formally be taken by a human.

The Court emphasised the importance of the GDPR's protection against decisions based exclusively on automated processing having legal or similarly significant effects.

Relevance

The case undermines the idea that an organisation can escape responsibility simply because:

“The algorithm made the decision.”

The legally accountable organisation remains subject to GDPR requirements.

Classification

Directly relevant to automated decision-making; not a case recognising AI moral personhood.

8. Österreichische Post – C-300/21

Österreichische Post AG v RW, C-300/21, CJEU, 4 May 2023

Principle

The CJEU considered compensation under GDPR Article 82 and confirmed that unlawful processing can give rise to compensation where the necessary requirements for damage and causal connection are satisfied.

The case is particularly relevant to AI because automated profiling may produce:

incorrect classifications;

reputational effects;

emotional harm;

discriminatory outcomes;

loss of control over personal data.

Artificial moral agency significance

The AI does not become the compensation debtor simply because it generated the profile.

The claim remains directed against the legally responsible controller or other responsible actor.

Classification

Directly relevant to automated profiling and AI accountability.

9. NAP – Case C-340/21

VB v Natsionalna agentsia za prihodite (NAP), C-340/21, CJEU, 14 December 2023

Principle

The case concerned a major personal-data security incident and the interpretation of GDPR security obligations and compensation.

The CJEU addressed questions concerning:

security measures;

risk assessment;

responsibility for data protection;

fear and non-material damage;

whether a data breach itself automatically establishes liability.

Relevance to artificial moral agency

AI systems frequently process enormous amounts of personal data.

If an AI system is inadequately secured, responsibility does not shift to the AI system itself. The relevant legal entity must demonstrate appropriate organisational and technical measures.

Classification

Directly relevant to AI/data-system accountability, but not specifically an AI moral-agent case.

10. Google Spain – Case C-131/12

Google Spain SL, Google Inc. v Agencia Española de Protección de Datos (AEPD), Mario Costeja González, C-131/12, CJEU, 13 May 2014

Principle

The CJEU recognised important data-protection obligations concerning search-engine processing and the right to request removal of certain results.

The case established that technologically mediated processing can create significant legal responsibilities for the operator.

Artificial moral agency significance

Google's search engine performs highly complex automated operations.

Nevertheless, the legal responsibility attaches to the operator/controller, rather than treating the algorithm itself as an autonomous legal person.

Broader lesson

Technological autonomy does not eliminate institutional responsibility.

Classification

Directly relevant to automated technological decision-making and data governance; analogical to moral agency.

11. Meta Platforms – Case C-252/21

Meta Platforms Ireland Ltd v Bundeskartellamt, C-252/21, CJEU, 4 July 2023

Principle

The CJEU examined the relationship between:

competition law;

personal-data processing;

consent;

contractual relationships;

behavioural advertising.

The case demonstrates how sophisticated algorithmic platforms may simultaneously be subject to several legal regimes.

Artificial moral agency significance

An algorithm may decide which advertisements, content or information users receive.

But the legal obligations remain attributable to the company operating the system.

The greater the platform's ability to design, configure and control algorithmic processing, the stronger the argument for organisational accountability.

Classification

Highly relevant analogical authority concerning algorithmic organisational responsibility.

12. CHEZ – Case C-83/14

CHEZ Razpredelenie Bulgaria AD v Komisia za zashtita ot diskriminatsia, C-83/14, CJEU, 16 July 2015

Principle

The CJEU considered indirect discrimination and the use of apparently neutral criteria that could disproportionately affect a protected group.

Relevance to AI

Modern AI systems frequently use:

geographical variables;

behavioural characteristics;

purchasing histories;

educational data;

employment histories;

credit indicators.

Such variables can operate as proxies for protected characteristics.

An algorithm therefore cannot necessarily avoid discrimination law merely because it does not expressly use race, sex, ethnicity or another protected characteristic.

Artificial moral agency significance

If an AI system generates discriminatory outcomes, the legal analysis focuses on the relevant organisation, decision process and discriminatory effect—not on attributing moral blame to the algorithm itself.

Classification

Analogical but highly important for algorithmic discrimination.

13. Feryn – Case C-54/07

Centrum voor gelijkheid van kansen en voor racismebestrijding v Firma Feryn NV, C-54/07, CJEU, 10 July 2008

Principle

The CJEU established that discriminatory recruitment statements can constitute evidence of discrimination even without identifying an individual victim who was actually rejected.

Relevance to AI

AI recruitment systems may reproduce:

discriminatory employer preferences;

historical workforce patterns;

biased training data;

discriminatory ranking criteria.

An employer cannot necessarily escape equality law by delegating recruitment decisions to software.

Artificial moral agency significance

The algorithm does not become the morally or legally responsible employer.

The organisation using the recruitment system remains legally relevant.

Classification

Analogical authority for algorithmic hiring discrimination.

14. Asociația Accept – Case C-81/12

Asociația Accept v Consiliul Național pentru Combaterea Discriminării, C-81/12, CJEU, 25 April 2013

Principle

The CJEU considered discriminatory statements and the evidential implications of statements suggesting discriminatory recruitment practices.

AI relevance

An AI recruitment model may appear neutral while reproducing discriminatory patterns.

Evidence may include:

model outputs;

statistical disparities;

training data;

evaluation records;

internal communications;

testing results.

Artificial moral agency significance

Again, the legal system generally asks:

Who created, deployed or relied upon the discriminatory system?

rather than:

Can the algorithm itself be punished?

Classification

Analogical authority.

15. Glukhin v Russia

Glukhin v Russia, Application No. 11519/20, ECtHR, 4 July 2023

Principle

The ECtHR considered the use of facial-recognition technology in connection with the applicant.

The Court examined the interaction between:

biometric processing;

privacy;

surveillance;

freedom of expression;

proportionality.

Artificial moral agency relevance

Facial-recognition technology can independently identify and classify individuals.

But technological autonomy does not transform the technology into a rights-bearing person.

The State remains responsible for the manner in which the system is deployed.

Classification

Highly relevant analogical ECtHR authority for AI surveillance and automated identification.

16. S. and Marper v United Kingdom

S. and Marper v United Kingdom, Applications Nos. 30562/04 and 30566/04, ECtHR Grand Chamber, 4 December 2008

Principle

The ECtHR examined retention of fingerprints and DNA profiles under Article 8 ECHR.

The case established important principles concerning:

personal data;

biometric information;

privacy;

proportionality;

State retention systems.

AI relevance

AI increasingly uses:

facial recognition;

voice recognition;

biometric identification;

behavioural profiling.

The case supports the proposition that technological processing of personal characteristics must remain subject to fundamental-rights safeguards.

Classification

Analogical authority.

17. Big Brother Watch and Others v United Kingdom

Big Brother Watch and Others v United Kingdom, Applications Nos. 58170/13, 62322/14 and 24960/15, ECtHR Grand Chamber, 25 May 2021

Principle

The ECtHR examined large-scale interception and surveillance regimes under Article 8 and Article 10 ECHR.

The Court emphasised the importance of safeguards surrounding powerful surveillance technologies.

AI relevance

Modern AI can dramatically increase the ability to:

analyse intercepted communications;

classify individuals;

identify behavioural patterns;

predict associations;

automate surveillance.

Artificial moral agency significance

The greater the autonomous analytical capacity of technology, the greater the importance of legal safeguards governing its human deployment.

Classification

Highly relevant analogical authority.

18. Delfi AS v Estonia

Delfi AS v Estonia, Application No. 64569/09, ECtHR Grand Chamber, 16 June 2015

Principle

The ECtHR examined platform responsibility for unlawful user-generated comments.

The case involved the tension between:

freedom of expression;

platform responsibility;

protection of others' rights.

AI relevance

Modern platforms increasingly use algorithms to:

recommend content;

moderate content;

rank comments;

suppress material;

identify potentially unlawful speech.

Artificial moral agency significance

Even where content moderation is substantially automated, the platform remains the relevant legal institution.

The algorithm does not ordinarily become the defendant.

Classification

Analogical authority concerning algorithmic/platform responsibility.

19. Glawischnig-Piesczek v Facebook Ireland – C-18/18

Glawischnig-Piesczek v Facebook Ireland Limited, C-18/18, CJEU, 3 October 2019

Principle

The CJEU considered injunctions concerning unlawful online content and the circumstances in which platforms may be required to identify and remove equivalent unlawful material.

Relevance to artificial moral agency

Content platforms increasingly rely upon automated systems for:

content identification;

filtering;

ranking;

moderation.

The case illustrates that automated processing can have legally significant consequences without making the algorithm itself the legal actor.

Classification

Directly relevant to automated platform governance; not a recognition of AI moral personhood.

20. Product Liability Analogy: Boston Scientific

Boston Scientific Medizintechnik GmbH and Others, Joined Cases C-503/13 and C-504/13, CJEU, 5 March 2015

Principle

The CJEU interpreted EU product-liability rules concerning defective medical devices.

The Court recognised that products can be defective where they fail to provide the safety that persons are entitled to expect.

AI relevance

The reasoning becomes important for AI-enabled products such as:

medical AI;

autonomous machines;

smart devices;

robotics;

AI-controlled industrial equipment.

If an AI-enabled product is defective, responsibility can arise under product-liability law.

Artificial moral agency significance

The product does not itself become the responsible legal person.

Liability is imposed through the applicable product-liability framework on the legally responsible economic actors.

Classification

Analogical/product-liability authority with significant AI relevance.

21. The Central European Legal Principle

The emerging European approach can be expressed as:

Autonomous operation does not necessarily mean autonomous legal responsibility.

An AI system can:

select;

rank;

predict;

classify;

recommend;

generate;

execute.

But the legal system asks who:

designed it;

supplied it;

trained it;

configured it;

deployed it;

supervised it;

benefited from it;

failed to monitor it;

failed to correct a known defect.

22. Can AI Be Sued Directly?

At present, generally no—not merely because it is autonomous.

A claimant normally needs a legally recognised defendant such as:

Private sector

developer;

manufacturer;

software provider;

platform;

employer;

insurer;

healthcare provider;

financial institution;

AI deployer.

Public sector

ministry;

municipality;

police authority;

immigration authority;

welfare authority;

public hospital;

educational institution.

The AI itself is generally treated as the technological instrument or system through which the legally relevant conduct occurred.

23. Can AI Be Criminally Responsible?

European criminal law presently does not generally treat an AI system as an autonomous criminal offender merely because it operates autonomously.

Criminal responsibility ordinarily requires a legally recognised offender and principles concerning:

actus reus;

mens rea;

intention;

recklessness;

negligence;

attribution;

legality;

culpability.

An AI may therefore create evidence of:

negligent programming;

reckless deployment;

failure to supervise;

corporate misconduct;

defective product design.

But that does not mean that the AI itself possesses criminal guilt.

24. AI Moral Agency and Human Oversight

Human oversight becomes particularly important when AI is capable of making consequential decisions.

A meaningful oversight system should include:

identification of high-risk decisions;

monitoring of model performance;

ability to intervene;

ability to stop the system;

review of unusual outputs;

documentation;

audit trails;

complaint procedures;

correction mechanisms;

appropriate expertise.

A person who merely clicks “approve” after an algorithm has made the substantive decision may not constitute meaningful human oversight.

25. Causation in Artificial Moral Agency Claims

Causation is often the most difficult issue.

Consider:

AI recommendation → human decision → harm

The claimant must determine whether:

the AI recommendation caused the decision;

the human independently exercised judgment;

the harm was foreseeable;

another factor caused the loss;

the AI output was merely advisory;

the decision-maker was required or expected to follow the AI recommendation.

The closer the human decision-maker was to merely rubber-stamping the AI output, the stronger the argument that the AI system was causally significant.

26. Foreseeability

Foreseeability is especially important in AI cases.

A defendant may be liable where it was reasonably foreseeable that an AI system would:

reproduce discriminatory data;

generate unsafe recommendations;

hallucinate important information;

misclassify people;

produce false positives;

leak personal data;

be vulnerable to adversarial attacks;

malfunction after deployment;

produce dangerous outputs.

However:

An unexpected AI error does not automatically establish negligence.

The claimant still normally needs to establish the applicable duty, breach, causation and legally recognised damage.

27. AI Moral Agency and Fundamental Rights

AI systems can affect rights protected by:

ECHR

Article 6 — fair trial;

Article 8 — private and family life;

Article 10 — expression;

Article 14 — non-discrimination;

Protocol No. 1 Article 1 — property.

EU Charter

Article 1 — human dignity;

Article 7 — private life;

Article 8 — data protection;

Article 11 — expression;

Article 16 — freedom to conduct a business;

Article 21 — non-discrimination;

Article 47 — effective judicial protection.

These rights belong to human beings and other legally recognised rights-holders, not automatically to the AI system making the decision.

28. Could AI Ever Become a Moral Agent Under European Law?

Theoretically, yes, but it would require a significant legal development.

Possible future models include:

Model 1 — Tool Model

AI remains a technological instrument.

Current European law largely follows this approach.

Model 2 — Limited AI Legal Status

AI could receive a narrowly defined legal status for specific functions without possessing general human rights.

Model 3 — Electronic Personhood

AI could theoretically become a special category of legal person.

This would require legislation defining:

rights;

duties;

assets;

liability;

representation;

registration;

insurance;

termination;

succession;

insolvency.

Model 4 — Full Moral Personhood

AI could theoretically receive rights comparable to those of persons.

This would require resolving extremely difficult questions concerning:

consciousness;

suffering;

autonomy;

identity;

copying;

deletion;

memory;

consent;

punishment;

equality.

There is presently no general European legal regime implementing this fourth model.

29. Why Legal Personhood Would Not Necessarily Solve AI Liability

Even if AI received some legal personality, a difficult question would remain:

Where would the money for compensation come from?

A legal person needs an economically meaningful liability structure.

Possible mechanisms might include:

mandatory insurance;

dedicated AI assets;

developer guarantees;

operator liability;

compensation funds;

pooled industry funds;

compulsory registration;

parental/creator liability.

Therefore, giving AI legal personality would not automatically solve the practical problem of compensation.

30. Case-Law Summary Table

31. Practical Legal Test for an Artificial Moral Agency Claim

A European court dealing with an AI-related harm can be analysed through the following sequence:

Step 1 — Identify the AI system

What did it actually do?

Step 2 — Identify the legal actor

Who developed, supplied, deployed or controlled it?

Step 3 — Identify the decision

Was the AI:

advisory;

partially automated;

fully automated;

merely executing a predetermined instruction?

Step 4 — Identify the legal regime

Potentially:

GDPR;

AI regulation;

equality law;

product liability;

consumer law;

contract;

tort/delict;

employment law;

administrative law;

competition law;

fundamental rights.

Step 5 — Establish the duty

What obligation applied to the relevant human/legal entity?

Step 6 — Establish breach

Was there:

defective design;

inadequate testing;

discriminatory training data;

inadequate supervision;

insufficient security;

failure to warn;

unlawful profiling;

lack of transparency?

Step 7 — Establish causation

Did the AI system materially contribute to the harm?

Step 8 — Establish damage

Possible damage includes:

physical injury;

economic loss;

privacy harm;

discrimination;

reputational damage;

psychological harm;

loss of opportunity;

unlawful exclusion.

Step 9 — Examine defences

Potential issues include:

contributory negligence;

unforeseeability;

independent human decision;

technical state of the art;

compliance with mandatory requirements;

third-party interference.

Step 10 — Select remedy

Possible remedies include:

compensation;

injunction;

correction;

deletion;

restriction of processing;

human reassessment;

reversal of decision;

access to information;

regulatory enforcement;

product recall;

declaratory relief.

32. Most Important Legal Distinction

The central distinction can be expressed as follows:

33. Conclusion

Artificial Moral Agency Claims are an emerging legal category rather than a recognised standalone European cause of action.

European law currently draws a critical distinction between technical autonomy and legal responsibility. An AI system may independently classify, recommend, predict, rank, generate or execute actions, but that does not presently make it an independent legal person or moral defendant.

The most important authorities—particularly SCHUFA, Österreichische Post, NAP, Google Spain, Meta Platforms, CHEZ, Feryn, Glukhin, S. and Marper, Big Brother Watch and Glawischnig-Piesczek—show a consistent direction: automated technology can produce legally significant consequences, but accountability is generally attributed to identifiable human or organisational actors.

Accordingly, the strongest present European formulation is:

AI may be functionally autonomous without being legally autonomous.

For a claimant, the practical objective is therefore not normally to prove that the AI itself is a moral person. Instead, the claimant should establish who designed, supplied, deployed, controlled, supervised or benefited from the AI system; what legal duty applied; how that duty was breached; how the AI contributed to the harm; and what remedy follows.

The concept could change in the future if European legislation were to recognise a special form of AI legal personality or electronic personhood, but such recognition would require major changes to European rules concerning rights, duties, liability, insurance, property, representation and judicial accountability.