Civil Law and AI-Controlled Space Mission Failure Accountability in Europe

1. Introduction

AI-controlled space mission failure accountability concerns situations where an artificial-intelligence system controls or materially assists a spacecraft, satellite, launch vehicle, rover, docking system, navigation system, autonomous guidance system, or mission-management process, and the mission fails.

Examples include:

AI incorrectly calculating a spacecraft trajectory;

autonomous navigation sending a satellite into an unsafe orbit;

AI failing to detect a collision risk;

an autonomous docking system damaging another spacecraft;

AI incorrectly managing propulsion or fuel;

an AI-controlled satellite losing communication;

software updates creating an unsafe autonomous decision;

training-data errors causing incorrect mission decisions;

human operators relying excessively on an AI recommendation;

cybersecurity manipulation of an autonomous space system.

There is not yet a large body of European case law directly deciding liability for an AI-controlled space mission failure. Therefore, European analysis has to combine space law, civil liability, product liability, contractual liability, non-contractual liability, AI regulation and general principles of causation and damage.

This distinction is important: several cases below are foundational or analogous authorities rather than cases actually involving an AI-controlled spacecraft.

2. Basic Legal Question

The central question is:

Who should compensate the victim when an AI-controlled space mission fails?

Potentially responsible parties include:

spacecraft/mission operator;

launch operator;

spacecraft manufacturer;

AI-system provider;

software developer;

satellite owner;

mission-control contractor;

human operator or supervisor;

infrastructure/service provider;

State or public authority;

insurer;

another spacecraft operator.

The fact that AI made the immediate decision does not normally mean that AI itself becomes the legal defendant.

The legal inquiry remains:

Duty → breach/defect → causation → damage → applicable liability regime → remedy.

3. European Legal Framework

Several legal regimes may operate simultaneously.

A. International space liability law

European space activities can be affected by:

Outer Space Treaty;

Liability Convention;

Registration Convention;

national space legislation;

licensing conditions;

insurance requirements;

contractual arrangements.

The Liability Convention is particularly important for damage caused by space objects, although the precise international responsibility regime is different from a private civil-liability claim between commercial parties.

B. EU AI law

The EU AI Act creates obligations concerning certain high-risk AI systems and AI used as safety components. Under the current framework, an AI system can qualify as high-risk where it is a safety component of a regulated product and the relevant conditions for third-party conformity assessment are satisfied. The AI Act also contains requirements concerning risk management, technical documentation, transparency and human oversight. (EUR-Lex)

The 2026 amendments further address AI systems whose failure or malfunction can endanger health and safety. (EUR-Lex)

However, the AI Act does not simply create a universal rule saying that an AI developer automatically pays for every AI-caused accident.

C. Product liability

A spacecraft is a physical product, but an AI-controlled mission can contain several layers:

spacecraft → embedded software → AI model → data → sensors → communications → human supervision.

European product-liability principles therefore become highly relevant.

The EU has also adopted the new Product Liability Directive, which modernises product liability for contemporary technologies and software.

D. Contract law

Space missions are usually based on complex contracts:

launch contracts;

satellite manufacturing contracts;

software-development agreements;

mission-operation agreements;

insurance contracts;

telecommunications agreements;

payload contracts;

service-level agreements.

A failure may therefore produce contractual claims independently of tort/product liability.

4. Main Types of AI Space-Mission Failure

4.1 Navigation failure

The AI calculates an incorrect trajectory.

Example:

AI predicts that a spacecraft can safely perform a manoeuvre, but the prediction is wrong and the spacecraft enters an unusable orbit.

Possible causes:

defective algorithm;

insufficient training data;

incorrect sensor data;

unexpected environmental conditions;

model drift;

inadequate validation.

4.2 Collision-avoidance failure

An autonomous system fails to identify another spacecraft or debris.

Potential claims may involve:

damage to spacecraft;

loss of mission;

loss of satellite services;

damage to another operator's property;

consequential economic loss.

4.3 Autonomous docking failure

AI controls docking and causes:

collision;

structural damage;

loss of cargo;

damage to another spacecraft.

The central issue becomes whether the failure resulted from:

AI defect + inadequate testing + defective sensors + operator failure + unforeseeable circumstances.

4.4 Propulsion-management failure

AI incorrectly manages:

fuel consumption;

thrust;

engine firing;

orbital correction.

A small algorithmic error may ultimately cause total mission loss.

4.5 Communication failure

AI may incorrectly:

prioritise communications;

shut down systems;

change communication frequencies;

interpret ground-control instructions;

isolate a subsystem.

The resulting loss may be especially difficult to investigate because the spacecraft itself may become inaccessible.

5. The Most Important Accountability Principle

AI does not automatically replace human legal responsibility

Suppose:

Company A owns spacecraft.
Company B supplies AI navigation software.
Company C operates mission control.
AI makes an incorrect manoeuvre.
Spacecraft is destroyed.

The court would not normally stop at:

“The AI made the decision.”

Instead it would ask:

Who designed the system?

Who trained it?

Who supplied the data?

Who validated it?

Who integrated it into the spacecraft?

Who authorised autonomous operation?

Who had the ability to override it?

Was human supervision adequate?

Were known failure modes identified?

Was the system used outside its intended purpose?

Was the failure reasonably foreseeable?

What caused the damage?

6. Case Law

Case 1 — Galileo International Technology and Others v Commission, T-279/03

This is one of the most relevant European cases because it directly concerned the Galileo satellite navigation project.

The applicants brought a damages action involving the Community's Galileo satellite global radio-navigation system. The General Court considered the principles governing non-contractual liability of the EU institutions, including the relationship between unlawful conduct, damage and causation. (Infocuria)

Importance

The case demonstrates that satellite-related activity can generate questions of:

non-contractual liability;

causation;

economic loss;

governmental/institutional conduct;

special damage.

Application to AI

If an EU institution, agency or public body were involved in an AI-controlled space programme, the claimant could potentially have to establish:

unlawful conduct + actual damage + causal connection.

It therefore provides a useful foundation for AI-space accountability.

Nature of authority: Directly space-related, but not an AI mission-failure case.

7. Case 2 — Airbus Defence and Space and Marlink Events v European Defence Agency, T-105/24

This is a particularly useful 2026 space-sector authority.

The General Court dealt with a European Defence Agency procurement concerning satellite communications, equipment and related services. The applicants alleged unlawful treatment in the procurement process and sought compensation for damage, including loss of opportunity. The Court addressed issues including:

obligation to state reasons;

compliance with procurement criteria;

manifest error of assessment;

equal treatment;

non-contractual liability;

compensation. (Curia)

Relevance to AI-controlled missions

Although the case did not concern an AI accident, it demonstrates how European courts can analyse disputes involving sophisticated satellite and space-related technological services through ordinary legal principles.

An AI-space dispute could similarly require the claimant to identify:

the legal obligation;

the specific technological decision;

the error;

causation;

economic damage.

Example

If an AI system used in selecting or managing a satellite-service provider makes an unlawful or erroneous decision, the claimant cannot simply prove that:

“The AI was wrong.”

The claimant must connect the AI-assisted decision to an actionable legal breach and compensable loss.

Nature of authority: Directly satellite-related, but procurement rather than mission-failure liability.

8. Case 3 — Boston Scientific Medizintechnik, Joined Cases C-503/13 and C-504/13

The CJEU considered defective medical devices under the EU Product Liability Directive.

The Court dealt with pacemakers and implantable cardioverter defibrillators that presented a potential safety defect. The Court accepted that products belonging to the same series or production group could be regarded as defective where there was an increased risk of failure, and replacement costs could form part of the damage necessary to restore the expected level of safety. (Infocuria)

Relevance to space AI

This is important for systemic AI defects.

Suppose an AI-controlled satellite system is found to have a serious design defect.

The operator may argue:

“Only one spacecraft actually failed.”

But product-liability analysis may ask whether the defect creates a sufficiently serious safety risk across a class of products.

Space example

If an AI guidance module installed in 50 satellites contains the same dangerous algorithmic defect, potential legal consequences could include:

investigation;

software correction;

replacement;

precautionary shutdown;

loss-of-use claims;

compensation.

Nature of authority: Analogical product-liability authority.

9. Case 4 — Dutrueux v Centre Hospitalier Universitaire de Besançon, C-495/10

The CJEU considered a national system under which a public hospital could be liable for damage caused by malfunctioning equipment even without fault on the part of the hospital.

The Court addressed the relationship between the EU Product Liability Directive and national liability regimes. (Infocuria)

Importance for space missions

This is highly relevant because a spacecraft failure may involve multiple overlapping liability regimes.

For example:

defective AI software → defective spacecraft → operator negligence → statutory space liability.

A claimant may therefore have more than one possible legal route.

The case illustrates that one must determine:

whether EU product liability applies;

whether another national liability regime applies;

whether the regimes are compatible;

whether the claimant is proceeding against the manufacturer or another operator.

Nature of authority: Analogical European product-liability authority.

10. Case 5 — Keskinäinen Vakuutusyhtiö Fennia v Koninklijke Philips, C-264/21

The CJEU considered who qualifies as a producer under the Product Liability Directive.

The dispute concerned a product bearing identifying characteristics of Philips. The Court interpreted the producer concept broadly enough to address a person who presents itself as the producer by putting its name, trademark or other distinguishing feature on the product. (Infocuria)

Application to AI space systems

Modern spacecraft can involve:

spacecraft manufacturer;

AI developer;

systems integrator;

branded satellite platform;

subcontractors.

The question may become:

Who legally presented itself as responsible for the technological product?

For example, if Company A markets a spacecraft as containing its proprietary autonomous navigation system, while Company B actually developed the AI, the allocation of producer responsibility may become significant.

Key principle

Commercial presentation and legal responsibility can matter, not merely who physically wrote the software.

Nature of authority: Analogical product-liability authority.

11. Case 6 — VI v KRONE – Verlag, C-65/20

This Austrian reference concerned inaccurate information published in a newspaper that allegedly caused physical injury.

The CJEU held that the inaccurate content of the newspaper did not itself make the newspaper a defective product under the EU Product Liability Directive. (Infocuria)

Importance for AI space systems

This case highlights a crucial distinction:

A defective physical product is not necessarily the same thing as defective information or intellectual content.

That distinction matters when an AI system gives an incorrect recommendation.

For example:

AI software gives a bad navigation recommendation;

spacecraft hardware is physically intact;

operator follows the recommendation;

spacecraft is lost.

The claimant may need to determine whether the claim is properly characterised as:

defective product;

defective software;

negligent professional service;

contractual breach;

failure of supervision;

ordinary tort liability.

Key lesson

The legal characterisation of the AI output can be decisive.

Nature of authority: Analogical authority concerning information and product liability.

12. Case 7 — Sanofi Pasteur, C-338/24

This is a significant recent CJEU authority from 26 March 2026 concerning defective-product liability.

The case concerned the interaction between:

product liability;

fault-based liability;

limitation periods;

progressive damage;

access to justice.

The CJEU considered when the limitation period begins in a situation involving progressive damage and addressed the relationship between the EU product-liability system and fault-based liability. (Infocuria)

Relevance to AI-controlled spacecraft

Space-mission damage may not always become apparent immediately.

Example:

AI incorrectly manages a satellite's battery system in 2027.
The degradation remains hidden.
The satellite finally fails in 2029.

A court may have to determine:

when the damage occurred;

when the claimant knew about it;

when the defect became discoverable;

which limitation period applies.

Importance

AI-space litigation may involve latent defects rather than instantaneous accidents.

Nature of authority: Analogical product-liability authority, particularly useful for limitation and latent damage.

13. Case 8 — Airbus Defence and Space and Marlink Events v European Defence Agency as a Causation Example

The 2026 General Court judgment also illustrates a broader point about loss of opportunity and compensation in sophisticated satellite-service disputes.

The case involved a claim for compensation arising from an allegedly unlawful procurement decision concerning satellite communications services. (Curia)

For AI-space litigation, this is useful when the claimant does not claim destruction of the spacecraft itself but instead alleges:

loss of a commercial opportunity;

lost mission contract;

lost satellite-service revenue;

additional mission expenses.

The claimant would still have to establish the necessary causal relationship between the actionable event and the claimed loss.

14. Who Can Be Liable?

A. Spacecraft operator

The operator may be liable where it:

deployed inadequately tested AI;

ignored warnings;

failed to supervise autonomous operations;

used AI outside approved parameters;

failed to update safety systems;

failed to maintain the spacecraft.

B. AI developer

Potential liability may arise where the AI developer:

supplied defective software;

failed to test known failure modes;

misrepresented system capabilities;

failed to provide necessary safety documentation;

supplied an inappropriate model;

failed to disclose known limitations.

C. Spacecraft manufacturer

The manufacturer may be implicated where:

AI is embedded in the spacecraft;

hardware and AI cannot realistically be separated;

sensors were defective;

system architecture was unsafe;

integration was defective.

D. Mission-control operator

Liability may arise if human controllers:

ignored AI warnings;

failed to intervene;

entered incorrect commands;

improperly configured autonomous mode.

E. Data provider

If the AI relies on external data, liability may potentially arise from:

incorrect orbital data;

inaccurate weather/space-weather information;

corrupted sensor information;

incorrect ephemeris;

outdated collision data.

But contractual terms and causation would be crucial.

15. AI Failure Versus Human Failure

A major issue is automation bias.

Suppose:

AI says “safe.”
Human controller has contrary warning information.
Controller follows AI.
Spacecraft is destroyed.

The question becomes:

Was the AI defective, or was the human decision negligent, or both?

A court may examine:

whether the AI recommendation was reasonably reliable;

whether the warning was understandable;

whether the operator was trained;

whether override mechanisms existed;

whether the operator had enough time to intervene.

16. The Black-Box Problem

AI-controlled spacecraft create an evidence problem.

After a mission failure, investigators may need:

AI evidence

model version;

model weights;

training data;

validation data;

system prompts/configuration;

safety constraints;

decision logs.

Spacecraft evidence

sensor data;

telemetry;

orbital position;

propulsion status;

communications;

temperature;

power levels.

Human evidence

operator commands;

override decisions;

mission-control communications;

emergency procedures;

training records.

Without such evidence, proving causation can be extremely difficult.

17. Causation Chain

A typical AI-space case can be analysed as:

Defective data

↓

AI produces incorrect prediction

↓

AI recommends unsafe manoeuvre

↓

Human/operator accepts recommendation

↓

Spacecraft performs manoeuvre

↓

Spacecraft enters unsafe condition

↓

Mission fails

↓

Economic/property/person-related damage

The claimant must identify where the legally relevant failure occurred.

18. Multiple Causes

Space missions are especially likely to involve concurrent causes.

Example:

AI prediction error = 40%
sensor malfunction = 20%
human error = 20%
communication delay = 10%
unexpected space-weather event = 10%

The actual legal analysis will not necessarily use these mathematical percentages.

Instead, the court may ask:

Was the AI error a substantial cause?

Was the damage foreseeable?

Was another event an intervening cause?

Did human intervention break the causal chain?

Did the operator accept an unreasonable risk?

19. Contractual Allocation of Risk

Space contracts are extremely important.

A contract may allocate responsibility for:

software defects;

system integration;

testing;

mission failure;

data accuracy;

cybersecurity;

consequential losses;

indemnification;

insurance;

force majeure;

liability caps.

However, contractual allocation cannot automatically eliminate every statutory or mandatory liability rule.

20. Product Liability and AI

The traditional product-liability model becomes complicated where the “product” is:

spacecraft + embedded software + AI model + continuously updated software.

The European Commission itself recognised the difficulty of applying traditional product-liability rules to AI and software and has pursued modernisation of the framework. (EUR-Lex)

This is particularly important for spacecraft because the AI may continue changing after launch.

21. Continuous-Learning AI

A particularly difficult situation is:

Spacecraft launched with Model A.
AI later receives an update.
It evolves to Model B.
Failure occurs after the update.

Potential defendants may argue:

original manufacturer supplied a safe system;

operator authorised the update;

software provider supplied the update;

new data caused the malfunction.

The court would therefore need to reconstruct the entire lifecycle of the AI system.

22. Cybersecurity and AI Mission Failure

Suppose hackers manipulate the AI's:

training data;

sensor data;

navigation instructions;

communication channels.

The legal analysis becomes more complicated.

Potential issues include:

cybersecurity negligence;

contractual security duties;

inadequate authentication;

failure to monitor;

unforeseeable third-party attack;

force majeure;

regulatory compliance.

The mere presence of a cyberattack does not automatically eliminate liability.

23. Damage That Can Be Claimed

Depending on the applicable legal regime, claims may include:

Property damage

spacecraft destruction;

satellite damage;

payload destruction;

damage to another space object.

Economic loss

loss of satellite services;

lost contracts;

lost revenue;

replacement launch;

emergency recovery costs.

Operational losses

mission delay;

additional fuel;

replacement spacecraft;

additional ground-control expenses.

Personal injury

If a space-mission failure ultimately causes injury or death on Earth or to persons involved in the mission, ordinary personal-injury principles may become relevant.

24. Importance of the Space Liability Convention

International space law must be considered separately from ordinary European civil liability.

A major distinction is:

International responsibility

The Liability Convention principally deals with responsibility between States for damage caused by space objects.

Private civil liability

A private company may instead bring proceedings under:

national civil law;

contract;

product liability;

insurance law;

applicable space legislation.

Therefore:

State international responsibility ≠ automatic private civil liability.

This distinction is fundamental in European space litigation.

25. AI Act and Space Missions

The AI Act becomes relevant where an AI system falls within its scope and classification rules.

For safety-related AI, the regulatory emphasis includes:

risk management;

technical documentation;

data governance;

monitoring;

human oversight;

accuracy;

robustness;

cybersecurity.

The current EU framework expressly recognises that an AI system whose failure or malfunction endangers health and safety can qualify as a safety component. (EUR-Lex)

However, whether a particular space AI system falls within a particular AI Act category must be determined from the actual product, use, jurisdictional scope and applicable Union legislation, rather than assuming that every space AI system is automatically “high risk.”

26. Evidence in AI Space-Mission Litigation

A claimant should ideally obtain:

AI model documentation;

software version history;

source-code records where legally obtainable;

telemetry;

sensor records;

orbital calculations;

mission-control logs;

operator communications;

AI recommendations;

human overrides;

cybersecurity records;

testing reports;

safety certification;

risk assessments;

maintenance records;

software-update records;

contractual specifications;

incident-investigation reports.

The most important evidence

Often it will be the AI decision log.

It should ideally show:

Input → AI processing → recommendation → confidence/uncertainty → safety checks → human intervention → final command.

27. Liability Matrix

This is not automatic liability; the claimant still has to establish the elements required by the applicable legal regime.

28. Defences

Potential defendants may argue:

1. Force majeure

The event was genuinely outside reasonable control.

2. Unforeseeable space environment

For example, an extraordinary event that could not reasonably have been anticipated.

3. Operator misuse

The AI system was used contrary to its instructions.

4. Human intervention

The AI gave a correct recommendation but a human entered the wrong command.

5. Third-party interference

A cyberattack or external spacecraft caused the failure.

6. Lack of causation

The AI error existed but did not cause the actual loss.

7. Contractual limitation

A valid contractual limitation may apply, subject to mandatory law.

29. Importance of Fault Allocation

The most difficult question may not be:

“Was the AI defective?”

It may instead be:

At which stage should the risk have been controlled?

For example:

Developer
→ should have tested algorithm.

Manufacturer
→ should have integrated it safely.

Operator
→ should have validated deployment.

Mission controller
→ should have monitored operation.

Regulator
→ may have imposed safety requirements.

Thus, liability may be distributed across the technological supply chain.

30. Six+ Case Laws at a Glance

The first six are particularly useful for constructing a doctrinal framework; the cases are not all direct precedents involving AI-controlled spacecraft. (Infocuria)

31. Practical Example

Assume a European company operates an autonomous Mars-probe system.

The AI incorrectly predicts the spacecraft's position.

It orders a trajectory correction.

The spacecraft misses the planned orbital insertion.

The mission fails and the company loses €500 million.

Step 1 — Identify the AI

Was the AI:

navigation software?

safety component?

decision-support tool?

autonomous control system?

Step 2 — Identify the defect

Was there:

algorithmic defect?

defective training data?

sensor failure?

software integration problem?

Step 3 — Identify responsibility

Possible defendants:

AI provider;

spacecraft manufacturer;

operator;

systems integrator.

Step 4 — Establish causation

Was the mission loss actually caused by the AI?

Step 5 — Determine damage

Was the claim for:

spacecraft destruction;

replacement costs;

lost revenue;

lost commercial opportunity?

Step 6 — Determine legal regime

Potentially:

space law + national civil law + contract + product liability + AI regulation + insurance.

32. Key Legal Principle

The strongest general principle for future European AI-space litigation is:

Autonomy does not eliminate accountability.

An AI system may make the operational decision, but legal responsibility must normally be assigned to the human or legal entities that:

designed;

supplied;

integrated;

deployed;

supervised;

maintained;

controlled

the system, depending on the applicable legal regime and the facts.

33. Exam-Oriented Answer Structure

For an examination or legal research paper, use this sequence:

AI-controlled space mission failure

↓

Identify spacecraft + AI system

↓

Identify operator/manufacturer/developer

↓

Determine applicable space-law regime

↓

Apply contract law

↓

Apply product liability

↓

Apply national tort/non-contractual liability

↓

Consider AI Act obligations

↓

Identify defect or breach

↓

Establish causation

↓

Determine damage

↓

Consider limitation/defences

↓

Determine compensation and other remedies

34. Ultra-Short Revision Notes

Remember these 10 points:

AI itself normally is not the legal defendant.

Space missions involve several overlapping liability regimes.

The operator may be responsible for deployment and supervision.

The AI developer may face claims concerning defective software.

The manufacturer may be liable for defective spacecraft or integrated systems.

Product-liability cases help analyse defective AI-enabled products.

Galileo International Technology is particularly useful for satellite-project and EU non-contractual liability principles. (Infocuria)

Boston Scientific is important for safety-related product defects. (Infocuria)

Fennia v Philips helps with identifying the legally relevant producer. (Infocuria)

The ultimate civil-law questions remain duty/breach or defect → causation → damage → remedy.

Conclusion

AI-controlled space mission failure accountability in Europe is an emerging area rather than a mature, AI-specific body of case law. The strongest present approach is to combine European space-law principles with product liability, contractual liability, national tort/non-contractual liability and the developing EU AI regulatory framework.

The key legal challenge is causal attribution: determining whether the mission failed because of the AI itself, defective data, hardware, software integration, human supervision, external interference, or a combination of these factors. European product-liability jurisprudence such as Boston Scientific, Dutrueux, Fennia, KRONE-Verlag and Sanofi Pasteur, together with space/satellite authorities such as Galileo International Technology and Airbus Defence and Space v EDA, provides a useful foundation for constructing that analysis. (Infocuria)