Civil Law And Fertility AI Embryo Selection Liability Claims In Europe

1. Introduction

Fertility AI embryo selection refers to the use of artificial intelligence, machine learning, image analysis, embryo time-lapse imaging, genetic information, or predictive algorithms to assist fertility specialists in deciding which IVF embryos should be selected for possible transfer.

The technology may be used to assess matters such as:

embryo morphology;

developmental patterns;

implantation potential;

chromosomal or genetic information;

likelihood of successful implantation;

risk of certain genetic conditions;

embryo viability.

This creates a new category of potential civil liability. A dispute could arise where an AI system:

incorrectly ranks embryos;

fails to identify a serious genetic abnormality;

wrongly classifies an embryo as unsuitable;

recommends an embryo that subsequently produces an affected pregnancy;

causes viable embryos to be discarded;

produces discriminatory or scientifically unsupported recommendations;

uses genetic data unlawfully;

gives fertility clinicians misleading information;

operates on biased or inadequately validated data;

makes a recommendation that the clinic treats as medically authoritative.

Important legal qualification: there is currently no major European court judgment directly deciding a civil damages claim against a fertility clinic or AI developer specifically for an AI embryo-selection error. The legal framework must therefore be constructed from existing IVF/embryo cases, medical-liability principles, data-protection law, product liability and AI regulation. The cases below are consequently identified as directly relevant IVF/embryo authorities or legal analogies, rather than being presented as cases that already decided AI-embryo-selection liability.

2. European Legal Framework

There is no single European civil-liability statute specifically regulating fertility AI.

The relevant framework can include:

A. National medical civil liability

National law generally determines:

medical negligence;

informed consent;

professional standards;

causation;

damages;

contractual liability;

limitation periods.

B. European Convention on Human Rights

Article 8 is particularly important because European Court of Human Rights jurisprudence recognises important aspects of reproductive autonomy and private life.

C. GDPR

Genetic and health information can constitute specially protected personal data. AI systems processing IVF/genetic information therefore raise:

lawful processing;

transparency;

data minimisation;

security;

automated decision-making;

data-subject rights.

D. EU AI Act

The AI Act, Regulation (EU) 2024/1689, is highly relevant to AI used in healthcare. Depending on the precise system and function, AI used as a medical device or safety-related healthcare system may fall within the Act's high-risk framework.

E. EU product liability

Directive (EU) 2024/2853 modernises European product liability and expressly accommodates software and AI-related products.

F. National assisted-reproduction legislation

Member States differ substantially regarding:

IVF;

embryo creation;

embryo storage;

preimplantation genetic testing;

embryo selection;

donor gametes;

embryo research;

disposal of embryos.

3. What Is the Legal Problem With AI Embryo Selection?

A traditional IVF decision may look like:

Embryologist → examines embryo → clinical judgment → embryo selected

AI-assisted IVF may instead look like:

Embryo images/genetic data
↓
AI algorithm
↓
Probability/ranking
↓
Embryologist/doctor
↓
Embryo selected
↓
Transfer
↓
Pregnancy/outcome

The central liability question becomes:

Who is legally responsible when the AI recommendation is wrong?

Potential defendants could include:

fertility clinic;

doctor;

embryologist;

AI developer;

medical-device manufacturer;

laboratory;

genetic-testing provider;

data processor;

distributor;

potentially another healthcare professional who relied on the output.

4. Case Law 1 — Costa and Pavan v Italy, Application No. 54270/10

This is the most directly relevant European human-rights case concerning genetic screening and embryo selection.

The applicants were healthy carriers of cystic fibrosis. They wished to use IVF together with preimplantation genetic diagnosis (PGD) to select an embryo unaffected by the disease.

Italian law prevented them from accessing the relevant procedure while permitting termination of a pregnancy where the foetus was affected by the same disease.

The ECtHR found a violation of Article 8, considering the Italian legal system internally inconsistent and the interference with the applicants' private and family life disproportionate. (HUDOC)

Principle

Reproductive decision-making, including access to techniques designed to avoid transmission of a serious genetic disease, falls within the sphere of private and family life.

Importance for AI embryo selection

The case does not establish a right to an AI-selected embryo or a right to a genetically healthy child.

Instead, it establishes an important starting point:

Regulation of embryo genetic selection can directly engage Article 8 rights.

Therefore, if an AI system is used to facilitate embryo selection, a legal dispute may involve not only negligence and damages but also questions of reproductive autonomy and consistency of the regulatory framework.

5. Case Law 2 — Evans v United Kingdom, Application No. 6339/05

Evans concerned embryos created through IVF using the applicant's eggs and her partner's sperm.

The applicant subsequently lost her ovaries because of cancer treatment, making the stored embryos extremely important to her possibility of having a genetically related child. Her former partner withdrew his consent to the embryos' continued use.

The Grand Chamber held that the UK's requirement of continuing mutual consent did not violate Article 8. The Court recognised that the dispute involved competing private-life interests and the interests of both genetic contributors. (HUDOC)

Principle

IVF embryos engage serious questions of:

reproductive autonomy;

consent;

private life;

competing genetic interests.

AI relevance

Suppose an AI system wrongly labels an embryo and a clinic destroys or refuses to preserve it.

Evans demonstrates why consent and control over IVF embryos cannot be treated simply as an ordinary commercial transaction.

An AI-selection dispute may therefore involve:

contract + medical negligence + reproductive autonomy + embryo regulation.

6. Case Law 3 — S.H. and Others v Austria, Application No. 57813/00

The applicants were infertile couples who wished to use assisted reproduction involving donor sperm or donor ova.

Austrian legislation restricted those techniques.

The Grand Chamber held that there was no violation of Article 8, while recognising that assisted reproduction falls within the sphere of private and family life. The Court also emphasised the sensitive ethical and scientific issues involved and afforded States a significant margin of appreciation. (HUDOC)

Principle

European States retain substantial regulatory discretion in sensitive assisted-reproduction questions because:

scientific developments are rapid;

ethical questions remain contested;

European legal systems differ.

AI significance

An AI embryo-selection system cannot automatically be assumed to be legally permissible throughout Europe merely because it is technologically possible.

A particular Member State may impose additional requirements concerning:

embryo selection;

genetic testing;

medical devices;

professional licensing;

embryo storage;

data processing.

7. Case Law 4 — Knecht v Romania, Application No. 10048/10

In Knecht, frozen embryos belonging to the applicant had been seized by authorities following concerns regarding the clinic holding them.

The applicant subsequently encountered difficulties in having the embryos transferred to a specialised clinic so that she could pursue IVF treatment.

The ECtHR accepted that Article 8 applied because private life encompasses decisions concerning whether and how to become a parent. It ultimately found no violation, but the case demonstrates the legal significance of governmental and institutional interference with stored embryos. (HUDOC)

AI significance

Suppose an AI system is connected to an embryo-management platform and an erroneous automated classification results in:

wrongful destruction;

inappropriate storage;

failure to transfer;

misidentification of an embryo.

Knecht is relevant to the seriousness of the underlying interest.

The embryo is not simply an ordinary item of commercial property.

8. Case Law 5 — Parrillo v Italy, Application No. 46470/11

The applicant had five embryos created through IVF with her deceased partner. She did not intend to proceed with pregnancy and wanted to donate the embryos for scientific research.

Italian law prohibited such donation.

The Grand Chamber held that Article 8 was applicable under the private-life aspect because the embryos contained her genetic material and represented an aspect of her identity. However, the Court found no violation, giving the State a wide margin of appreciation on the sensitive issue of embryo research. (HUDOC)

Principle

Embryos created through IVF can engage significant private-life interests, but the Convention does not establish a uniform European legal regime for every question concerning embryos.

AI significance

AI embryo-selection companies may generate large amounts of:

embryo images;

genetic profiles;

developmental records;

patient information.

Parrillo demonstrates why the legal status and treatment of embryos cannot be reduced to ordinary commercial data or ordinary property.

9. Case Law 6 — Brüstle v Greenpeace, C-34/10

The CJEU considered the meaning of "human embryo" under EU biotechnology patent law.

The Court adopted a broad interpretation for the purposes of Directive 98/44/EC and held that patentability could be excluded where an invention required the use or destruction of human embryos. The Court linked the interpretation to respect for human dignity. (Eur-Lex)

Principle

EU law may treat the human embryo as raising distinctive issues of:

human dignity;

biotechnology;

public policy;

ethical regulation.

AI significance

Brüstle is not a civil-liability case and does not decide whether AI embryo-selection software is lawful.

Its importance is conceptual:

European law may impose special restrictions on technological activities involving human embryos because ordinary commercial-property concepts are insufficient.

This becomes relevant if AI companies seek to commercialise technologies involving embryo data, embryo manipulation or embryo-derived biological material.

10. Case Law 7 — Boston Scientific Medizintechnik GmbH, Joined Cases C-503/13 and C-504/13

This CJEU case concerned defective medical devices.

The Court held that where products belonging to the same production group or series present an unusually high risk of failure, the relevant defect may be established at the group level without proving that every individual product has independently manifested the same failure.

Principle

Product liability can address systemic risks rather than requiring proof of an identical physical defect in every individual product.

AI embryo-selection application

Imagine an AI embryo-selection medical device is trained using defective methodology and systematically produces unreliable embryo rankings.

If the system belongs to a class of products with a demonstrated safety or reliability problem, Boston Scientific provides an analogy for arguing that the problem may be systemic rather than merely an isolated bad outcome.

However, the analogy must be used carefully because embryo-selection AI presents different scientific and legal questions from implanted medical devices.

11. Case Law 8 — Sanofi Pasteur MSD SNC v Centre Hospitalier Universitaire de la Réunion, C-621/15

This case concerned alleged injury from a vaccine and addressed proof of causation under EU product-liability law.

The CJEU recognised that national courts may, subject to EU-law requirements, consider a body of evidence when assessing causation where scientific evidence does not provide straightforward proof.

AI significance

AI embryo-selection disputes can involve extremely complex causation.

For example:

AI recommendation
↓
embryologist relies on ranking
↓
embryo selected
↓
pregnancy
↓
genetic/medical outcome

A claimant may need to establish:

the AI output was defective;

the clinic relied upon it;

a different reasonable decision would probably have been made;

the different decision would probably have altered the outcome;

legally recognised damage resulted.

That is substantially more difficult than merely proving that the AI prediction was inaccurate.

12. The Most Important Liability Question: What Is "Damage"?

A major difficulty is that an incorrect embryo-selection decision does not automatically equal legally compensable damage.

Consider three situations.

Situation A — Wrongly rejected embryo

AI says:

Embryo A has low implantation potential.

The clinic chooses Embryo B.

Embryo B fails to implant.

Can the parents claim that Embryo A would have succeeded?

That requires difficult proof of counterfactual causation.

Situation B — AI fails to identify genetic abnormality

AI recommends an embryo that later results in a pregnancy affected by a serious genetic condition.

Potential claims could involve:

negligent testing;

failure to warn;

inadequate validation;

professional negligence;

wrongful birth-type claims where recognised by national law.

But the precise availability and measure of damages differs substantially between European jurisdictions.

Situation C — AI incorrectly destroys an embryo

This may raise a different category of claim involving:

reproductive autonomy;

breach of contract;

negligence;

emotional harm;

loss of reproductive opportunity;

Article 8 rights.

Evans, Knecht and Parrillo become particularly relevant to the legal significance of the embryo and the reproductive decision. (HUDOC)

13. Wrongful Selection vs Wrongful Birth

These concepts should not be confused.

Wrongful selection

The complaint is:

"The clinic/AI selected the wrong embryo."

Wrongful birth

The complaint may be:

"Because of negligent medical information or testing, the parents were deprived of the opportunity to avoid a particular pregnancy."

European jurisdictions differ significantly concerning whether and how such claims are recognised.

Therefore, a European-wide statement such as:

"Parents automatically receive damages when AI fails to detect a genetic condition"

would be legally incorrect.

14. AI Developer Liability

An AI developer could potentially face liability where the system itself is defective.

Possible allegations:

inadequate training;

inadequate validation;

foreseeable bias;

incorrect probability calculations;

failure to disclose limitations;

defective software;

inadequate cybersecurity;

inadequate updates;

misleading claims about accuracy.

Under the modern EU product-liability framework, software can fall within the concept of a product.

This is particularly important for fertility AI because the software may perform a medically significant function even though it has no physical components.

15. Fertility Clinic Liability

The clinic is likely to be central to most practical disputes because the AI recommendation is generally integrated into clinical decision-making.

Potential allegations include:

Failure to validate

The clinic adopts an AI system without adequate evidence that it performs reliably on the relevant patient population.

Blind reliance

The doctor treats the AI score as conclusive despite its intended role as decision support.

Failure to supervise

The clinic does not maintain appropriate human oversight.

Failure to inform

Patients are not adequately informed that AI contributed to embryo selection.

Failure to document

The clinic cannot reconstruct:

what data were entered;

what algorithmic output was produced;

what version of the model was used;

what clinical decision followed.

16. Doctor or Embryologist Liability

A healthcare professional may remain responsible for professional judgment even where AI is involved.

A potential legal argument would be:

"The AI made the recommendation, but the clinician made the medical decision."

This creates an important distinction between:

AI error

and

clinical reliance on AI error.

A doctor may potentially be liable where a reasonable professional should have recognised that:

the AI output was unreliable;

the patient's circumstances fell outside the model's validated population;

the system produced an obviously anomalous result;

confirmatory testing was necessary.

The exact professional standard is determined principally by national law.

17. Human Oversight

For fertility AI, human oversight is particularly important.

A legally safer structure is:

AI analysis
↓
Embryologist reviews result
↓
Medical assessment
↓
Patient receives appropriate information
↓
Final clinical decision

The presence of a human does not automatically eliminate liability.

If the human simply approves every AI recommendation without meaningful review, the existence of nominal human oversight may not resolve the underlying negligence question.

18. Genetic Data Liability

Embryo-selection AI can process exceptionally sensitive information.

Potential data include:

DNA information;

genetic disease indicators;

family medical history;

reproductive information;

embryo images;

health records.

Genetic and health data receive heightened protection under the GDPR.

A dispute could therefore contain two separate causes of action:

Medical claim

The AI/clinic selected an embryo negligently.

Data-protection claim

The clinic or AI provider unlawfully processed or disclosed genetic information.

These should not be conflated.

19. Automated Decision-Making

GDPR Article 22 can become relevant where a decision based solely on automated processing produces legal or similarly significant effects.

However, not every AI-assisted clinical recommendation automatically constitutes an Article 22 decision.

The legal analysis depends upon:

whether the decision is solely automated;

whether human intervention is meaningful;

the nature of the decision;

the effect on the individual;

the applicable legal basis.

Therefore:

AI-assisted medical decision ≠ automatically prohibited automated decision.

20. AI Act and Fertility Systems

The EU AI Act creates a regulatory layer separate from ordinary civil liability.

Depending on the system's precise function and regulatory classification, an AI system used in fertility medicine may be subject to requirements concerning:

risk management;

data governance;

technical documentation;

record keeping;

transparency;

human oversight;

accuracy;

robustness;

cybersecurity.

The important distinction is:

Regulatory breach ≠ automatically civil liability.

But evidence of regulatory non-compliance may be relevant to a national negligence or product-liability claim.

21. Product Liability for Fertility AI

A simplified future claim could be:

Defective AI software
↓
Unreasonable embryo recommendation
↓
Clinic relies upon recommendation
↓
Embryo transferred
↓
Legally recognised damage
↓
Product-liability claim

The claimant would still have to satisfy the applicable requirements concerning:

defect;

damage;

causation.

The new EU Product Liability Directive is therefore highly relevant to the future development of this field.

22. The Causation Problem

This is probably the hardest issue.

Assume:

AI ranked Embryo A: 80% predicted implantation;

Embryo B: 60%;

clinic selected B;

B failed;

A was never implanted.

Can the parents prove that A would have resulted in a successful pregnancy?

Not necessarily.

Embryo implantation is affected by multiple factors.

Therefore, the claimant must distinguish:

AI prediction error

from

legally sufficient causation.

A prediction can be scientifically wrong without automatically causing compensable damage.

23. Loss of Chance

Some national legal systems recognise certain forms of loss-of-chance reasoning, while others impose stricter causation requirements.

In fertility litigation this could be particularly important.

For example:

AI incorrectly rejected an embryo, reducing the patient's opportunity for successful implantation.

The legal question could become whether the claimant lost a legally recognisable chance rather than whether successful birth was certain.

This issue cannot be given one uniform European answer because national damages law differs.

24. Wrongful Disposal of Embryos

A particularly serious scenario is:

AI system incorrectly classifies an embryo as non-viable and the embryo is discarded.

Potential legal issues include:

negligence;

breach of fertility-treatment contract;

informed consent;

reproductive autonomy;

emotional damage;

Article 8;

national embryo-protection legislation.

Evans is especially relevant because it demonstrates the profound reproductive interests associated with stored IVF embryos. (HUDOC)

Parrillo likewise confirms that embryos may engage significant private-life interests, although the Court gave States a wide margin in the particular context of research donation. (HUDOC)

25. AI Bias in Embryo Selection

An AI model could perform differently for different populations because its training data may not adequately represent:

ethnic/genetic diversity;

age groups;

fertility conditions;

donor populations;

different IVF protocols.

A biased system could potentially result in systematic misclassification.

Possible legal routes include:

medical negligence;

product liability;

discrimination law;

data-protection law;

AI regulatory law.

However, whether a particular disparity constitutes legally actionable discrimination depends on the facts and applicable national/EU law.

26. Informed Consent

A patient may argue:

"I was never told that an algorithm would participate in selecting my embryo."

This raises a distinction between:

Consent to IVF

and

Consent/information concerning AI-assisted decision-making.

Whether disclosure of AI involvement is legally required, and the consequences of inadequate disclosure, depend upon applicable medical and data-protection rules.

Nevertheless, transparency becomes increasingly important where the AI system materially influences a reproductive decision.

27. Contractual Liability of Fertility Clinics

A fertility clinic may have contractual obligations concerning:

IVF services;

laboratory standards;

embryo storage;

genetic testing;

selection procedures;

confidentiality;

informed consent.

If the clinic promises a particular validated genetic-testing procedure but uses a substantially different or inadequately validated AI system, contractual liability may arise independently of strict product liability.

28. Manufacturer vs Clinic: A Practical Allocation

The actual defendant depends on the contractual and technical architecture.

29. Case-Law Comparison

30. Six Core Legal Principles

Principle 1 — Reproductive choices receive significant legal protection

Costa and Pavan establishes the importance of access to reproductive techniques involving genetic screening. (HUDOC)

Principle 2 — Consent is fundamental

Evans demonstrates that consent to IVF can remain legally significant throughout the treatment process. (HUDOC)

Principle 3 — States retain regulatory discretion

S.H. shows that European States retain considerable discretion concerning sensitive assisted-reproduction regulation. (HUDOC)

Principle 4 — Embryos are legally sensitive objects of regulation

Knecht and Parrillo illustrate the special legal significance of IVF embryos. (HUDOC)

Principle 5 — Human dignity matters in embryo-related biotechnology

Brüstle demonstrates the relevance of human dignity in EU biotechnology law. (Eur-Lex)

Principle 6 — Technical products can create systemic liability

Boston Scientific provides an important analogy for AI systems that produce systematic safety risks.

31. Future Liability Scenario

Consider this hypothetical:

A fertility clinic purchases an AI embryo-selection system.

The manufacturer claims:

"95% accuracy in identifying embryos with high implantation potential."

The system was trained predominantly on data from one demographic population.

A patient undergoes IVF.

The system ranks an embryo as unsuitable.

The clinic follows the recommendation.

The embryo is not transferred.

A later independent assessment indicates that the AI's classification was unreliable.

The patient brings proceedings.

Potential claims

Against AI manufacturer

defective software;

misleading performance claims;

inadequate validation;

defective algorithm.

Against clinic

inadequate professional judgment;

inappropriate reliance on AI;

inadequate patient information;

failure to validate the system.

Under data law

unlawful processing;

inadequate transparency;

improper use of genetic information.

Under human-rights law

potential Article 8 issues depending upon State involvement and the particular facts.

But the claimant must still establish a legally recognised injury and causal connection.

32. Major Difficulty: "Wrong Embryo" Does Not Automatically Mean "Defective AI"

This distinction is essential.

An embryo may fail to implant even when the AI correctly classified it.

Likewise, an embryo predicted to have high implantation potential may fail.

Therefore:

AI prediction probability is not a guarantee of pregnancy.

A court would need to consider:

algorithm accuracy;

validation methodology;

patient-specific circumstances;

competing clinical factors;

embryology evidence;

alternative treatment decisions;

statistical probability;

medical causation.

33. Civil-Law Remedies

Depending on national law, potential remedies may include:

Damages

For recognised:

physical injury;

psychological injury;

medical expenses;

reproductive harm;

other legally recognised losses.

Contractual remedies

refund;

treatment costs;

compensation;

termination of service agreement.

Injunction

To prevent continued processing or use of unlawfully obtained genetic data.

Data remedies

access;

correction;

deletion where legally available;

restriction of processing;

compensation for GDPR violations where applicable.

Regulatory remedies

investigation;

suspension;

correction;

withdrawal of system;

enforcement against provider.

34. Why European Civil Law Is Particularly Complex

A single AI embryo-selection dispute can involve:

Medical law
↓
Contract law
↓
Tort law
↓
Product liability
↓
GDPR
↓
AI Act
↓
Genetic-data regulation
↓
Embryo law
↓
Human rights
↓
EU/national procedural law

Consequently, there is unlikely to be a single cause of action capable of resolving every dispute.

35. Direct Cases vs Analogical Cases

This distinction is very important for legal research.

Directly relevant IVF/embryo cases

Costa and Pavan v Italy

Evans v United Kingdom

S.H. and Others v Austria

Knecht v Romania

Parrillo v Italy

These cases directly concern IVF, embryos, reproductive autonomy or embryo-related regulation.

Analogical technology/product cases

Brüstle v Greenpeace

Boston Scientific

Sanofi Pasteur

These do not decide AI embryo-selection liability. They provide principles concerning embryo biotechnology, defective products and medical causation.

This distinction should be maintained in any litigation memorandum.

36. Conclusion

Fertility AI embryo-selection liability in Europe is an emerging field rather than a settled independent branch of European civil law.

The central legal structure is:

AI system → clinical recommendation → embryo selection → reproductive outcome → alleged damage

The principal liability questions are:

Was the AI system defective or scientifically unreliable?

Was the clinic negligent in selecting or relying upon it?

Was the patient adequately informed?

Was genetic/health data processed lawfully?

Who controlled the algorithm and its updates?

Was the system properly validated for the relevant population?

Did the AI error actually cause legally compensable damage?

What national law governs the damages claim?

Were reproductive-autonomy or Article 8 interests affected?

Does EU product-liability law apply to the particular software/system?

The existing European case law provides a strong foundation but does not yet establish a specific rule that an AI embryo-selection error automatically creates civil liability. The most important existing authorities are Costa and Pavan on PGD and reproductive autonomy, Evans on IVF consent, S.H. on assisted-reproduction regulation, Knecht and Parrillo on the legal significance of embryos, and Brüstle, Boston Scientific and Sanofi Pasteur for biotechnology, product-defect and causation analogies. (HUDOC)

Ultra-short revision formula

Fertility AI Liability =

AI Embryo Analysis
→ Genetic/Medical Data
→ AI Ranking/Selection
→ Clinical Reliance
→ Embryo Transfer or Rejection
→ Reproductive Outcome
→ Defect + Duty + Causation + Damage
→ Clinic / Doctor / AI Provider / Manufacturer
→ GDPR + AI Act + Product Liability + National Medical/Tort Law + Article 8 ECHR.