Civil Law And AI-Controlled Space Mission Decision Accountability Claims In Europe

1. Introduction

AI-controlled space mission decision accountability concerns civil and public-law responsibility when an AI system makes, recommends, or materially influences decisions during a space mission.

Examples include AI deciding or recommending:

orbital manoeuvres;

collision-avoidance actions;

launch or abort decisions;

satellite station-keeping;

autonomous navigation;

spacecraft attitude control;

communication-routing decisions;

robotic landing;

resource allocation;

emergency responses;

spacecraft shutdown;

rendezvous or docking;

planetary-surface operations.

A typical claim might look like:

An autonomous spacecraft-control AI incorrectly predicts a collision trajectory, performs an unnecessary manoeuvre, collides with another satellite, and causes economic loss.

Or:

An AI-controlled spacecraft fails to execute an emergency avoidance manoeuvre, resulting in the destruction of another space object.

The legal difficulty is that AI may make the operational decision, but European law generally does not treat the AI itself as the ultimate legal bearer of responsibility. Responsibility must instead be allocated among the spacecraft operator, owner, manufacturer, AI provider, launch entity, State, EU institution, contractor and other legally responsible actors.

There is currently no reported CJEU or ECtHR judgment directly deciding a damages claim arising from an AI-controlled autonomous space mission. The legal framework must therefore be constructed from:

European AI law;

EU product liability;

European space-programme law;

international space liability;

general civil/tort principles;

existing European satellite cases; and

case law concerning automated decisions and hazardous activities.

2. Basic Accountability Principle

The central principle can be stated as:

Autonomous operation does not automatically mean autonomous legal responsibility.

If an AI system makes the operational choice:

AI → decision → spacecraft action → damage

the legal inquiry normally continues:

Who designed it? → Who deployed it? → Who controlled it? → Who could intervene? → Was the system defective? → Was the decision foreseeable? → Who suffered damage? → Which liability regime applies?

3. European Legal Framework

The relevant legal regimes are particularly complex because space activities operate simultaneously at several levels.

Level 1 — International space law

Important instruments include:

Outer Space Treaty 1967;

Liability Convention 1972;

Registration Convention 1975;

Rescue Agreement 1968.

The Liability Convention establishes an international State-centred framework for damage caused by space objects. For damage on the surface of the Earth, Article II establishes a strict-liability model; for damage occurring elsewhere than on the surface of the Earth, Article III applies a fault-based rule between launching States. The Convention is therefore not simply an ordinary private tort regime.

Level 2 — EU law

Important instruments include:

EU Space Programme Regulation 2021/696;

EU AI Act 2024/1689;

Product Liability Directive 2024/2853;

GDPR where personal data are involved;

EU Charter of Fundamental Rights.

Regulation 2021/696 establishes the EU Space Programme and includes Galileo, EGNOS, Copernicus, space surveillance and tracking, and GOVSATCOM. (EUR-Lex)

Level 3 — National space law

Member States may impose:

licensing requirements;

insurance requirements;

operator liability;

safety requirements;

indemnification rules;

registration obligations.

Level 4 — General civil law

National law can additionally govern:

negligence;

contract;

product liability;

property damage;

economic loss;

contribution between defendants;

indemnification.

4. AI Act and Autonomous Space Systems

The AI Act is relevant where the AI system falls within its territorial and material scope.

A particularly important provision is Article 6.

The current consolidated AI Act treats an AI system as high-risk where it is a safety component of a product, or is itself a product, covered by specified Union harmonisation legislation and the relevant conformity-assessment conditions are satisfied. The amended text also states that an AI system whose failure or malfunction would endanger health and safety qualifies as a safety component for the relevant analysis. (EUR-Lex)

However, an important qualification is necessary:

Not every spacecraft AI automatically becomes a high-risk AI system merely because it operates in space.

The precise classification depends upon the AI Act's scope and the applicable product legislation.

5. AI Risk Management

Where the AI Act applies to a high-risk system, its risk-management approach is highly relevant.

The system must be subjected to an ongoing process of:

identifying risks;

analysing risks;

evaluating risks;

adopting mitigation measures;

testing;

monitoring;

reassessing risks.

The AI Act's framework is particularly relevant to autonomous systems because the risk does not necessarily arise only from initial programming.

AI can change through:

retraining;

updates;

new operational data;

model adaptation;

environmental changes.

Thus:

A spacecraft AI can become legally problematic even if its original design was reasonable.

6. Product Liability Directive 2024/2853

This is one of the most important developments for AI-controlled spacecraft.

The new Product Liability Directive expressly expands the concept of a product to include software and AI systems.

It states that software can be a product whether supplied:

independently;

through a device;

through a network;

through cloud technology;

through software-as-a-service.

It also treats AI-system providers as manufacturers for the relevant product-liability framework. (EUR-Lex)

The Directive applies to products placed on the market or put into service after 8 December 2026. (EUR-Lex)

Importance for spacecraft

Suppose:

Autonomous navigation AI + spacecraft hardware

contains a defect that causes a collision.

The AI may not simply be treated as an intangible service. Depending on the circumstances and applicability of the Directive, software can fall within the product-liability regime.

7. Continuing Liability After Launch

The new Product Liability Directive is particularly important for autonomous systems because it recognises that manufacturers can retain control over products after they have been placed on the market.

The Directive expressly discusses:

software updates;

upgrades;

machine-learning algorithms;

cybersecurity vulnerabilities.

It can therefore cover defectiveness arising after initial market placement where relevant software or related services remain under the manufacturer's control. (EUR-Lex)

This is highly significant for autonomous spacecraft.

Imagine:

2027: AI passes testing.

2029: manufacturer supplies an update.

2030: updated AI causes an incorrect orbital manoeuvre.

The legal analysis cannot necessarily stop at:

“The spacecraft was safe when originally launched.”

8. Case 1 — Airbus Defence and Space and Marlink Events v European Defence Agency, T-105/24

Court: General Court of the European Union
Judgment: 1 July 2026

This is currently one of the most directly relevant recent European space-sector decisions.

The dispute concerned a public procurement procedure involving satellite communications, equipment and related services.

The General Court found, among other things, errors in the evaluation process and annulled the relevant procurement decisions. It also awarded compensation for loss of opportunity. (Curia)

The Court awarded:

EUR 3,864,315 to Airbus Defence and Space;

EUR 458,185 to Marlink Events,

for the loss of opportunity associated with the procurement decision. (Curia)

Why it matters for AI-controlled missions

This is not an AI-liability case and should not be presented as one.

Its relevance is that it demonstrates that European courts can deal with:

satellite communications;

institutional decision-making;

manifest errors of assessment;

reasoning obligations;

equal treatment;

non-contractual liability;

loss of opportunity;

compensation.

These concepts can become relevant if an AI system is used in a European space procurement or operational decision and produces an actionable institutional error.

Current procedural status

An appeal, C-1004/26 P, was lodged before the Court of Justice in September 2026 and remains pending. (Infocuria)

9. Case 2 — Galileo International Technology and Others v Commission, T-279/03

Court: General Court
Judgment: 10 May 2006

This case concerned the EU's Galileo satellite navigation project and a claim for non-contractual liability against the Community.

The claimants alleged damage connected with use of the name “Galileo” by the EU satellite project.

The General Court considered:

non-contractual EU liability;

damage;

unlawful conduct;

unusual and special damage;

the legal consequences of the Galileo project.

The action was ultimately dismissed. (Infocuria)

AI relevance

Again, this is not an AI case.

Its significance is that it is genuine European case law concerning:

EU satellite infrastructure + non-contractual liability.

It demonstrates that an EU space programme can generate disputes concerning compensation and institutional responsibility.

For an AI-controlled Galileo-type system, questions could arise regarding whether damage resulted from:

EU institutional action;

contractor conduct;

satellite operator conduct;

AI-system defect;

operational negligence.

10. Case 3 — KF v European Union Satellite Centre, C-464/20 P

Court: CJEU
Judgment: 14 October 2021

This case concerned the European Union Satellite Centre (SatCen) and a dispute involving its staff and an administrative investigation.

The CJEU dismissed the appeal. (Infocuria)

Importance

This case is useful because it demonstrates that the EU Satellite Centre exists within a legal accountability framework rather than operating outside ordinary EU judicial review.

For AI-controlled satellite operations, responsibility could therefore involve:

EU institutional law;

employment law;

administrative decision-making;

damages;

judicial review.

Again, it is analogical rather than direct AI-mission authority.

11. Case 4 — SCHUFA Holding, C-634/21

Court: CJEU
Judgment: 7 December 2023

This is a major European automated-decision case.

The CJEU held that an automated probability value can fall within Article 22 GDPR where it plays a determining role in a subsequent decision producing significant effects. (Infocuria)

Application to space missions

Imagine an autonomous spacecraft generates:

“Collision probability: 96%.”

The mission controller then automatically authorises an emergency manoeuvre.

If the probability was generated from personal data, Article 22 may be relevant in a GDPR context.

But even where GDPR does not apply, the case illustrates a broader legal concept:

A numerical AI prediction can become legally significant when it effectively determines a consequential decision.

For autonomous spacecraft, the equivalent problem is:

AI probability → operational decision → physical consequence.

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

Court: CJEU
Judgment: 27 February 2025

The CJEU addressed the right to receive meaningful information concerning the logic involved in automated profiling and decision-making. (Curia)

The Court's reasoning is especially relevant to autonomous systems because accountability requires some ability to understand how the system arrived at its output.

Space-mission application

Suppose an AI decides:

“Abort landing.”

The mission operator asks:

“Why?”

A useful accountability framework would require sufficient information to determine:

which sensor information was used;

which risk variables mattered;

whether the AI detected an anomaly;

whether the model had conflicting inputs;

whether a software update changed the decision;

whether the system operated within its design parameters.

This does not necessarily mean disclosure of all source code.

It means that an unexplained autonomous decision may create serious difficulties for:

causation;

fault;

auditability;

regulatory compliance;

litigation.

The case therefore provides an important explainability analogy. (Curia)

13. Case 6 — Boston Scientific Medizintechnik, Joined C-503/13 and C-504/13

Court: CJEU
Judgment: 5 March 2015

Boston Scientific concerned defective medical devices.

The CJEU held, in substance, that where products belonging to the same group or production series have a potential defect creating an abnormal risk, individual products may be treated as defective even where the defect has not been individually demonstrated in each particular device. (Infocuria)

AI-controlled spacecraft application

Suppose:

30 spacecraft use the same autonomous-navigation AI;

telemetry reveals that the AI has a systematic trajectory-selection defect;

one spacecraft has not yet malfunctioned.

A claimant could argue by analogy that the systemic defect is relevant even before an individual mission experiences the same failure.

This is particularly important for:

satellite constellations;

reusable spacecraft;

identical spacecraft platforms;

common AI software;

common autonomous-navigation modules.

Limitation

Boston Scientific is a medical-device product-liability case, not a space case.

Its value is doctrinal:

A systemic safety defect may be legally significant even when an individual unit's precise failure cannot yet be isolated.

14. Case 7 — W and Others, C-621/15

Court: CJEU
Judgment: 21 June 2017

W and Others concerned proof of defect and causation under the EU product-liability regime where scientific consensus was lacking.

The CJEU accepted that, subject to the conditions established in its judgment, serious, specific and consistent evidence could be relevant to proving defect and causal connection where scientific evidence did not provide a definitive answer. (curia)

Importance for autonomous spacecraft

This is potentially very important.

Imagine:

AI executes a manoeuvre → spacecraft is lost.

But telemetry is incomplete.

The operator says:

“We cannot prove that the AI caused the accident.”

The claimant may rely on:

mission logs;

repeated similar failures;

temporal proximity;

software version;

system alerts;

other spacecraft exhibiting similar behaviour;

engineering evidence.

W and Others provides a useful analogy for dealing with scientific and technical uncertainty in causation.

15. Case 8 — Köbler v Republic of Austria, C-224/01

Court: CJEU
Judgment: 30 September 2003

Köbler established the principle that Member States may be liable for damage caused by sufficiently serious infringements of EU law attributable even to courts of last instance, subject to the applicable conditions. (Infocuria)

Space-AI relevance

Consider an EU institutional space decision:

AI-generated analysis → EU institution adopts unlawful decision → individual/company suffers damage.

Köbler is not directly a space or AI case, but it demonstrates a broader European principle:

The fact that a decision is produced within a public institutional system does not automatically exclude State/Union liability where the relevant legal conditions are satisfied.

16. Case 9 — Traghetti del Mediterraneo, C-173/03

Court: CJEU
Judgment: 13 June 2006

Traghetti concerned non-contractual State liability for damage caused by an infringement of EU law attributable to a national court of last instance.

The Court rejected a blanket exclusion of State liability merely because the alleged error concerned legal interpretation or assessment of facts/evidence. (Infocuria)

AI relevance

It provides an important accountability principle:

Technical or interpretive complexity does not automatically eliminate legal responsibility.

Applied analogically to autonomous spacecraft:

An operator cannot necessarily defend a claim simply by saying:

“The AI made the decision, so no human actor can be responsible.”

The legal system must examine:

who controlled the system;

who was responsible for deployment;

whether the error was foreseeable;

whether safeguards were required;

whether the system was defective;

whether intervention was possible.

17. Case 10 — Budayeva and Others v Russia

Court: ECtHR
Judgment: 20 March 2008

Budayeva concerned deaths caused by a mudslide and the State's failure to take adequate measures against a foreseeable natural hazard.

The ECtHR's case law recognised positive obligations concerning foreseeable risks and the need for appropriate preventive measures. (ECHR-KS)

Space-AI application

The analogy becomes particularly interesting for:

space-debris risks;

uncontrolled re-entry;

collision avoidance;

foreseeable spacecraft failures;

emergency warnings.

Suppose a European space operator knows that an autonomous AI system has a recurring tendency to misclassify debris.

The operator nevertheless continues using it without mitigation.

Budayeva supports the broader legal concept that foreseeable serious risks can generate preventive obligations.

Limitation

Budayeva is a natural-disaster/human-rights case, not a private AI or space-liability case.

18. Case 11 — Öneryıldız v Turkey

Court: ECtHR Grand Chamber
Judgment: 30 November 2004

Öneryıldız concerned deaths caused by a dangerous activity and the State's obligations relating to protection of life and regulation of hazardous activities.

The case is important for the European doctrine that authorities may have positive obligations to regulate dangerous activities and take reasonable preventive measures.

Space-AI relevance

Space missions can involve extraordinary risks:

high-energy launch systems;

orbital collisions;

uncontrolled re-entry;

nuclear or radioactive power sources in certain missions;

hazardous propulsion systems;

large satellite constellations.

If an autonomous system controls such activities, the legal analysis should not assume that automation reduces the duty to manage foreseeable risks.

19. The Special Problem of Autonomous Decision-Making

The most important distinction is between:

1. Human-commanded mission

Human makes decision → spacecraft executes.

2. AI-assisted mission

AI recommends → human approves → spacecraft executes.

3. Human-supervised autonomous mission

AI decides → human can intervene → spacecraft executes.

4. Fully autonomous mission

AI decides → no immediate human intervention → spacecraft executes.

The more operational authority is transferred to AI, the more important the following become:

design safeguards;

validation;

testing;

intervention mechanisms;

logging;

monitoring;

emergency shutdown;

explainability;

cybersecurity.

20. Can the AI Itself Be Sued?

Generally, AI is not treated as an independent legal person merely because it makes autonomous decisions.

The claim normally has to identify a legally responsible actor such as:

spacecraft owner;

operator;

manufacturer;

AI developer;

AI provider;

launch provider;

mission contractor;

insurer;

State;

EU institution.

Thus:

Autonomous decision ≠ autonomous legal personality.

21. Operator Liability

The spacecraft operator is often the most obvious actor for investigation.

Questions include:

Did the operator properly test the AI?

Did it know of previous failures?

Was the AI within approved operational parameters?

Was human intervention available?

Were emergency protocols established?

Were software updates monitored?

Were collision warnings ignored?

Did the operator knowingly deploy an immature model?

The operator's precise liability will depend on applicable international and national law.

22. Manufacturer Liability

A spacecraft manufacturer may potentially face liability where:

hardware was defective;

software was defective;

AI integration was defective;

safety controls were inadequate;

an update caused the malfunction;

the system failed to meet legally relevant safety expectations.

The new Product Liability Directive is particularly important because it expressly includes software and AI within the concept of products. (EUR-Lex)

23. AI Provider Liability

An independent AI provider might be responsible where its system:

contains a design defect;

generates foreseeable dangerous outputs;

lacks adequate testing;

receives inadequate safety constraints;

has defective updates;

has cybersecurity weaknesses;

produces erroneous autonomous recommendations.

The contractual allocation between the spacecraft operator and AI provider will therefore be extremely important.

24. Continuous-Learning AI

This is one of the most difficult issues.

Suppose:

Day 1: AI is safe.

Day 100: AI receives new operational data.

Day 200: AI changes its decision policy.

Day 201: spacecraft crashes.

Who is responsible?

Potentially relevant questions include:

Who authorised learning?

Who controlled retraining?

Was the updated model validated?

Was the change documented?

Was human approval required?

Could the operator revert to the previous model?

Was model drift foreseeable?

The Product Liability Directive is especially relevant because it recognises software and machine-learning-related changes under the manufacturer's control as potentially relevant to defectiveness. (EUR-Lex)

25. AI Hallucination in Mission Operations

A spacecraft AI could theoretically produce an erroneous interpretation such as:

“Telemetry indicates engine failure.”

when the engine is actually functioning normally.

If the AI then commands shutdown, the resulting damage could be enormous.

This creates several possible claims:

defective software;

negligent integration;

inadequate validation;

inadequate human oversight;

contractual breach;

product liability;

operational negligence.

The key evidence would include:

input data;

model output;

confidence level;

model version;

logs;

operator instructions;

previous warnings;

safety constraints.

26. AI and Collision Avoidance

Collision avoidance is perhaps the clearest hypothetical.

Example

Satellite A and Satellite B approach one another.

AI predicts:

Collision probability = 3%.

No manoeuvre occurs.

Actual collision occurs.

A second system would have predicted:

Collision probability = 72%.

The legal dispute would then involve:

A. Was the AI defective?

B. Was the input data defective?

C. Was the prediction within expected error margins?

D. Was the operator required to use another verification system?

E. Was the AI appropriately validated?

F. Did the operator ignore an available warning?

G. Was the collision attributable to another spacecraft?

H. Which international liability regime applies?

27. Causation

AI-space claims will probably face extremely complicated causation questions.

A spacecraft loss may involve:

AI error + sensor failure + software update + space weather + communication delay + operator instruction + hardware malfunction.

Therefore, courts may need to determine:

Was the AI error a necessary cause, substantial contributing cause, or merely one background factor?

The reasoning in W and Others is useful by analogy where scientific certainty is unavailable. (Infocuria)

28. Evidence and AI Audit Logs

AI-controlled missions create a new category of evidence.

Important records may include:

model version;

training-data version;

telemetry;

sensor readings;

decision logs;

confidence scores;

software updates;

operator interventions;

cybersecurity events;

communications with ground control;

emergency commands.

Without these records, establishing causation may be extremely difficult.

This makes traceability a major component of accountability.

29. Explainability

An autonomous mission decision should ideally be reconstructable.

For example:

Input: radar detects object
↓
AI calculation: collision probability 83%
↓
Decision: avoidance manoeuvre
↓
Constraint: fuel reserve
↓
Final command: 4.5° orbital adjustment

This type of record can allow a court to determine:

whether the AI acted within its design;

whether the decision was reasonable;

whether the data were correct;

whether a human could intervene;

whether the AI was defective.

Dun & Bradstreet provides an important European analogy for the importance of meaningful information about automated decision logic. (Curia)

30. Cybersecurity Liability

An autonomous spacecraft can also be attacked.

Suppose:

hacker alters navigation data → AI processes false information → spacecraft performs dangerous manoeuvre.

The question becomes:

Was the cyberattack alone responsible, or was the system inadequately secured?

The Product Liability Directive specifically recognises the relevance of software updates necessary to address cybersecurity vulnerabilities when analysing continuing product defectiveness. (EUR-Lex)

Therefore:

AI error + cybersecurity vulnerability

may create a different liability analysis from:

AI error + unforeseeable external cyberattack.

31. Space-Debris Accountability

Autonomous spacecraft make space-debris management especially important.

A defective AI may:

fail to avoid debris;

generate debris through collision;

incorrectly de-orbit a satellite;

miscalculate re-entry;

interfere with another spacecraft.

The EU's space-surveillance framework already recognises the importance of monitoring space infrastructure, orbital objects and uncontrolled re-entry risks. EU space policy has expressly contemplated collision-risk and liability-related uses of space-surveillance information. (EUR-Lex)

32. Damage on Earth vs Damage in Space

This distinction is crucial under international space law.

Damage on Earth's surface

The Liability Convention provides a stronger strict-liability structure for the launching State.

Example:

AI-controlled satellite re-enters unexpectedly and damages a building in France.

Damage elsewhere than Earth's surface

For example:

AI-controlled satellite collides with another satellite in orbit.

The international Liability Convention uses a fault-based approach between launching States for such damage.

Therefore:

The location of the damage can materially affect the applicable international liability regime.

This international space-law framework is separate from possible private claims under national civil law.

33. Multiple Defendants

A single autonomous mission accident may involve:

Spacecraft manufacturer
↓
AI developer
↓
Sensor manufacturer
↓
Launch provider
↓
Spacecraft operator
↓
Mission controller
↓
Government/space agency

The claimant may therefore face a complex multi-party causation problem.

Contribution and indemnity rules may become as important as the initial liability question.

34. Contractual Allocation

Space missions are often heavily contractual.

Contracts may allocate responsibility for:

software defects;

updates;

mission control;

data quality;

AI performance;

insurance;

indemnification;

limitation of liability;

force majeure;

cybersecurity;

emergency intervention.

Therefore, two parties could agree internally that:

AI provider indemnifies operator for specified software defects.

But such a contract does not necessarily determine the rights of third-party victims or override mandatory law.

35. Loss of Opportunity

The 2026 Airbus Defence and Space and Marlink Events v EDA judgment is useful because it demonstrates that European Union courts can recognise loss of opportunity as a compensable category in an appropriate institutional context. (Curia)

In a future AI-space dispute, analogous questions could arise concerning:

lost mission opportunity;

lost launch window;

lost satellite capacity;

lost commercial contracts;

lost scientific mission opportunity.

However, the claimant would still need to satisfy the applicable causation and damages requirements.

36. Important Distinction: Regulatory Breach vs Civil Liability

This is essential for examination purposes.

Suppose an AI operator violates an AI Act obligation.

That does not automatically mean:

every person affected is entitled to civil damages.

Conversely, a claimant might have a civil claim even where the specific AI Act provision does not itself provide a damages action.

The legal analysis should therefore separate:

Regulatory compliance

Was the AI Act complied with?

Civil liability

Was there a legally actionable breach causing damage?

Contract

Was a contractual obligation breached?

Product liability

Was the product defective?

Space law

Which international liability regime applies?

37. Major Case-Law Table

The first three are the strongest space-sector authorities; the remainder supply the legal doctrines that may govern an AI-space dispute.

38. Potential Claim Structure

A future claimant might plead:

Claim 1 — Defective AI

The autonomous decision system was defective.

Claim 2 — Negligent deployment

The operator deployed an insufficiently tested system.

Claim 3 — Failure to monitor

Known model degradation was ignored.

Claim 4 — Failure to intervene

Human intervention was available but not used.

Claim 5 — Contractual breach

The AI provider failed to meet agreed performance or safety obligations.

Claim 6 — Product liability

The AI/software/spacecraft constituted a defective product under applicable law.

Claim 7 — Cybersecurity negligence

The system was insufficiently protected.

Claim 8 — Regulatory breach

Applicable AI/space-sector requirements were violated.

Claim 9 — International space liability

The relevant launching State may bear liability under the Liability Convention.

39. Possible Defences

The defendant might argue:

1. The AI performed within specifications

No defect existed.

2. Unforeseeable space event

The event could not reasonably have been predicted.

3. Independent third-party fault

Another spacecraft caused the collision.

4. Sensor failure

The AI received incorrect information.

5. Operator override

The human operator caused the accident.

6. Cyberattack

The system was manipulated by an external actor.

7. Contractual limitation

The claimant's loss falls outside the contractual risk allocation.

8. Lack of causation

The AI error did not actually cause the damage.

These are factual and legal issues to be established under the applicable regime rather than automatic defences.

40. Future Problem: AI as the “Mission Commander”

The most difficult scenario would be:

No human makes the immediate operational decision.

The spacecraft itself:

interprets sensor data;

predicts danger;

evaluates alternatives;

chooses the manoeuvre;

executes the command.

This raises the philosophical appearance of an “AI decision-maker,” but legally the important question remains:

Which human or legal entity assumed responsibility for giving the AI that authority?

That could be:

spacecraft operator;

space agency;

commercial company;

manufacturer;

mission contractor.

41. Legal Accountability Chain

The most useful model is:

AI MODEL

↓

SPACECRAFT SOFTWARE

↓

SPACECRAFT HARDWARE

↓

MISSION OPERATOR

↓

GROUND CONTROL

↓

SPACE AGENCY / COMPANY

↓

STATE / INTERNATIONAL LIABILITY

This prevents the common analytical mistake of treating the AI as if it were automatically the defendant.

42. Key Legal Formula

For a civil claim:

AI DECISION → DEFECT/BREACH → HUMAN/ENTITY RESPONSIBILITY → CAUSATION → SPACE DAMAGE → LEGALLY RECOGNISED LOSS → REMEDY

For product liability:

DEFECTIVE AI/SOFTWARE → DAMAGE → CAUSATION → PRODUCER/ECONOMIC OPERATOR LIABILITY

For international space liability:

SPACE OBJECT → LAUNCHING STATE → DAMAGE → APPLICABLE LIABILITY CONVENTION → COMPENSATION

For autonomous decision-making:

INPUT → AI PROCESSING → PREDICTION → AUTONOMOUS DECISION → ACTION → DAMAGE → AUDIT TRAIL

43. Exam-Oriented Conclusion

AI-controlled space mission decision accountability in Europe is an emerging field with no mature body of direct case law specifically dealing with autonomous AI spacecraft decisions.

The existing legal framework nevertheless provides substantial tools.

The EU Space Programme Regulation establishes the institutional framework for major European space systems such as Galileo, EGNOS, Copernicus, space surveillance and GOVSATCOM. (EUR-Lex)

Airbus Defence and Space and Marlink Events v EDA, T-105/24 is an important 2026 satellite-sector authority demonstrating that errors in institutional decision-making concerning satellite communications can lead to annulment and non-contractual compensation, including compensation for loss of opportunity. (Curia)

Galileo International Technology, T-279/03 confirms that EU satellite projects can generate disputes concerning non-contractual liability. (Infocuria)

SCHUFA and Dun & Bradstreet provide important principles concerning automated decision-making and explainability. (Infocuria)

Boston Scientific and W and Others provide useful product-liability principles concerning systemic defects and difficult questions of causation. (Infocuria)

The new Product Liability Directive 2024/2853 is particularly significant because it expressly brings software and AI systems within the EU product-liability framework, including certain defects arising through controlled updates and machine-learning-related changes. It applies to products placed on the market or put into service after 8 December 2026. (EUR-Lex)

Ultimately, AI autonomy does not automatically eliminate legal accountability. The central legal task is to identify the entity that designed, supplied, deployed, controlled or failed to supervise the autonomous system and then establish defect or breach, causation and legally recoverable damage.

Ultra-basic revision formula

AI → SPACECRAFT → DECISION → ACTION → DAMAGE → OPERATOR/PROVIDER/STATE → CAUSATION → LIABILITY → REMEDY

One-line exam answer

“In European law, an AI-controlled spacecraft may make an autonomous operational decision, but responsibility ordinarily remains attributable to the legally responsible operator, manufacturer, provider, institution or State under the applicable civil, product-liability and space-law regimes.”