Civil Law and Comparative AI Maritime Regulation Models

Detailed legal analysis with comparative frameworks and case laws

1. Introduction

The relationship between civil law, artificial intelligence (AI), and maritime regulation is becoming increasingly important as shipping industries adopt autonomous vessels, AI-assisted navigation, predictive maintenance, automated cargo handling, smart ports, and algorithmic collision-avoidance systems.

Maritime law traditionally allocates legal responsibility among shipowners, shipmasters, charterers, cargo interests, manufacturers, insurers, and other parties. AI complicates this framework because an accident may result from a combination of human error, software defects, faulty sensor data, cybersecurity failures, or decisions made by an autonomous system.

Civil law is particularly relevant because it provides rules governing compensation for injury, property damage, contractual breaches, defective products, and negligence. Comparative legal analysis helps identify how different jurisdictions may allocate liability when AI systems operate in international waters.

Central argument: Existing maritime liability principles remain essential to regulating AI-enabled shipping, but effective regulation requires clearer rules concerning human oversight, autonomous-system certification, evidence and data preservation, insurance, product liability, and the allocation of responsibility among multiple actors.

This analysis examines the major legal concepts, compares leading regulatory models, and discusses more than six relevant judicial decisions. It distinguishes established case law from proposed AI-specific legal solutions.

2. Meaning and scope of the principal concepts

2.1 Civil law

The expression civil law has two meanings in this context:

Civil-law legal tradition: A legal system based substantially on codified legislation, such as those found in France, Germany, and many other European and Asian jurisdictions.

Civil liability: The legal obligation to compensate another person for injury, loss, damage, or breach of a legal duty. Civil liability exists in both civil-law and common-law jurisdictions.

This distinction is important because maritime AI regulation concerns both legal methodology and the substantive rules governing compensation.

Civil liability commonly arises through:

Contractual liability: Breach of a carriage contract, charterparty, shipbuilding contract, software licence, or maintenance agreement.

Tort or non-contractual liability: Negligence, unlawful conduct, or breach of a duty of care causing damage.

Product liability: Injury or loss caused by a defective autonomous-navigation system, sensor, or other product.

Vicarious or organizational liability: Liability attributed to an employer, shipowner, operator, or other legally responsible entity for acts within the relevant legal relationship.

Statutory liability: Responsibility imposed by maritime safety legislation, pollution conventions, or other applicable laws.

2.2 AI in maritime operations

Maritime AI refers to computational systems capable of performing or supporting tasks such as perception, prediction, classification, planning, and operational decision-making.

Checking Favorites status

AI-assisted navigation

Collision-risk prediction, route optimization, radar interpretation, and navigational decision support.

Checking Favorites status

Autonomous and remotely operated vessels

Ships that perform some or all navigational functions with reduced or no continuous onboard human intervention.

Checking Favorites status

AI-enabled ports and cargo systems

Automated cranes, cargo inspection, berth scheduling, predictive maintenance, and logistics management.

The International Maritime Organization (IMO) has used the concept of Maritime Autonomous Surface Ships (MASS) for ships that can operate independently of human interaction to varying degrees. Autonomy is not a single technical condition: a vessel may be remotely controlled, partly autonomous, or highly autonomous.

2.3 Maritime regulation

Maritime regulation is the combination of international conventions, domestic legislation, classification and certification requirements, judicial decisions, contractual rules, and administrative standards governing ships and shipping activities.

Its principal objectives are:

Safety of life at sea.

Prevention of collisions and maritime pollution.

Protection of cargo, passengers, seafarers, and third parties.

Allocation of liability and compensation.

Seaworthiness and technical reliability.

Cybersecurity and operational resilience.

Effective investigation and enforcement.

AI regulation must address both ex ante obligations—testing, certification, risk assessment, and operational safeguards—and ex post liability, including compensation and remedies after an accident.

3. The legal problem: who is responsible when AI causes maritime damage?

Consider an autonomous cargo vessel that fails to detect another ship and causes a collision.

The immediate cause might be a defective camera, an inadequately trained model, poor integration of radar and navigation data, an erroneous software update, an unsafe deployment decision, or a failure by a remote operator to intervene.

The legal analysis should therefore separate technical causation from legal responsibility.

Liability does not automatically attach to every party listed. It depends on the applicable law, the evidence, the relevant duties, causation, contractual terms, and any statutory limits or defences.

A central difficulty is that AI systems may be technically complex and partly opaque. The party suffering damage may lack access to training data, software logs, system updates, or the model's operational records. Consequently, evidentiary rules and preservation of digital records can be as important as the substantive law of negligence.

4. International legal framework governing AI maritime operations

Maritime AI regulation is not governed by a single universal AI liability convention. Instead, several international instruments interact with national civil liability law.

4.1 United Nations Convention on the Law of the Sea (UNCLOS), 1982

UNCLOS establishes the broader international legal framework for maritime jurisdiction, navigation, flag-state responsibilities, and protection of the marine environment.

Its relevance to AI maritime systems includes:

Flag-state jurisdiction: Ships are generally subject to the jurisdiction and regulatory responsibilities of their flag State.

Navigation and safety: Autonomous ships must operate consistently with applicable international navigation obligations.

Marine environmental protection: AI systems used for pollution prevention must comply with applicable environmental duties.

Jurisdictional complexity: A collision may involve ships of different flags, damage in another State's waters, and software supplied from a third country.

UNCLOS does not itself establish a complete system of civil liability for every AI-related collision. Applicable collision conventions, domestic legislation, and private international law may determine the substantive claim and the competent court.

4.2 International Convention for the Safety of Life at Sea (SOLAS), 1974

SOLAS is a major international treaty concerning ship safety. Its rules address construction, equipment, operation, navigation, and other safety matters.

For AI-enabled vessels, relevant questions include whether:

Navigation and control systems meet applicable safety requirements.

Remote-operation facilities are reliable.

The ship can operate safely under foreseeable conditions.

Emergency procedures remain effective when autonomous systems fail.

Human personnel have appropriate training and intervention capabilities.

Compliance with technical safety standards is important evidence, but it does not necessarily eliminate civil liability. A vessel can comply with minimum regulatory requirements and still be operated negligently in particular circumstances.

4.3 International Regulations for Preventing Collisions at Sea (COLREGs), 1972

COLREGs govern collision avoidance, including lookout, safe speed, assessment of collision risk, and appropriate manoeuvres.

The central issue is whether a collision-avoidance system can perform functions that the rules expect from a competent human navigator.

For example, an AI system should not merely identify another vessel. It must assess collision risk, account for uncertainty, interpret the applicable navigation rules, and take timely and appropriate action.

The legal difficulty arises when a system's technical decision conflicts with the conduct expected of a competent mariner. Investigators must examine both the machine's behaviour and the decisions made by the persons responsible for its operation.

4.4 The International Safety Management (ISM) Code

The ISM Code emphasizes safe management, operational procedures, emergency preparedness, and systematic risk control.

For AI shipping, an effective safety management system should address:

Software validation before deployment.

Version control and authorized software updates.

Cybersecurity and protection against data manipulation.

Human intervention and remote-control arrangements.

Recording and reporting system failures.

Contingency planning for sensor or communications outages.

An operator that deploys an AI system without assessing its foreseeable failure modes may face allegations of inadequate risk management, even where the immediate malfunction originates in software developed by another company.

4.5 The IMO Maritime Autonomous Surface Ships Code: position as of October 2026

An important recent development is the IMO's adoption of the non-mandatory International Code of Safety for Maritime Autonomous Surface Ships (MASS Code) in May 2026, effective from 1 July 2026. The Code provides a goal-based framework addressing autonomous-vessel safety, operational risk, certification, cybersecurity, and human oversight. The IMO's stated roadmap envisages further work towards a mandatory code, with adoption expected by 1 July 2030 and entry into force on 1 January 2032. These are planned milestones, not an already completed mandatory international regime.

Autonomous shipping

+2

The Code is significant because it moves regulation beyond simply asking whether a ship has a human crew. It considers the functions performed by the ship, the conditions in which autonomy is permitted, and the arrangements for monitoring and responding to failures.

However, a safety code and a civil liability regime serve different purposes. A code may establish how an autonomous ship should operate; a liability regime determines who must compensate an injured person when the system causes damage.

5. Comparative AI maritime regulation models

Different jurisdictions approach AI-related maritime risks through different combinations of codified law, judicial precedent, safety regulation, product liability, and administrative oversight.

Checking Favorites status

Model 1 — European civil-law and risk-regulation model

France, Germany and the wider EU framework

Codified civil liability, structured product-safety rules, risk-based AI governance, and regulatory supervision.

Checking Favorites status

Model 2 — English common-law maritime model

United Kingdom

Judicial precedent, negligence, contractual allocation of risk, seaworthiness, and established maritime conventions.

Checking Favorites status

Model 3 — United States liability and regulatory model

Tort litigation, product liability, federal maritime law, statutory obligations, and substantial reliance on case-specific adjudication.

Checking Favorites status

Model 4 — Singapore's technology and maritime-governance model

Singapore

Maritime administration, structured innovation trials, port regulation, operational safety, and commercial dispute resolution.

Checking Favorites status

Model 5 — India's hybrid maritime model

Codified statutes, constitutional and statutory regulatory powers, judicial precedent, and international maritime obligations implemented through domestic law.

5.1 European civil-law and risk-regulation model

The European approach combines national civil codes with EU-level rules on product safety, AI, and liability.

In France, the Civil Code distinguishes contractual responsibility from liability arising outside a contract. In Germany, the Civil Code (Bürgerliches Gesetzbuch or BGB) provides the framework for fault-based damages, while separate legislation addresses particular forms of product liability.

The EU's AI Act uses a risk-based regulatory structure, imposing more demanding requirements on specified high-risk AI uses. Whether a particular maritime AI application falls within a category regulated as high-risk must be assessed under the Act's actual scope and classifications; not every AI system used on a ship automatically qualifies.

The EU's revised Product Liability Directive (Directive (EU) 2024/2853) modernizes product-liability rules to address contemporary products, including software and certain digital components. Its applicability depends on the Directive's scope and national implementation, including transitional rules.

Civil liability implications

A software defect may be examined as a product defect, where the applicable product-liability rules permit.

A shipowner or operator may face separate negligence or contractual claims.

Product-liability and negligence claims may coexist, subject to rules against double recovery and other applicable limitations.

Technical records and access to evidence can be critical to proving a defect and causal connection.

Strengths: Structured legislation, relatively clear categories of liability, and a strong emphasis on product safety and risk prevention.

Weaknesses: Potential overlap between general civil liability, product liability, AI-specific obligations, and maritime conventions. Different legal bases may apply to different actors in the same accident.

5.2 English common-law maritime model

English law relies heavily on judicial precedent, supplemented by legislation and international maritime conventions.

Important legal doctrines include negligence, contractual interpretation, seaworthiness, causation, remoteness of damage, and the allocation of risk under charterparties and bills of lading.

For AI shipping, the key question is often whether a shipowner, operator, manufacturer, or contractual party acted reasonably and complied with its legal obligations in the circumstances.

A software defect does not automatically make the shipowner liable for every consequence. The court must determine the applicable duty, the relevant breach, the causal relationship, and any contractual or statutory allocation of risk.

Strengths: Flexible judicial reasoning, developed maritime jurisprudence, and sophisticated treatment of contractual risk allocation.

Weaknesses: AI-specific rules may be less directly codified, and litigation can be expensive when software failures involve multiple jurisdictions and technically complex evidence.

5.3 United States model

The United States combines federal maritime law, statutory regulation, and state-law doctrines where applicable.

Potential claims may involve maritime negligence, product liability, breach of warranty, and statutory obligations. The precise legal route depends on jurisdiction, the maritime connection, the parties, and the governing law.

For an AI-enabled vessel, a claimant might allege that:

A manufacturer supplied defective navigation software.

A shipowner failed to maintain or safely configure the system.

An operator failed to respond to warnings.

A company failed to address foreseeable cybersecurity risks.

These are distinct allegations. A court must evaluate the legal basis of each claim rather than assuming that one malfunction establishes liability for every participant.

Strengths: Established litigation mechanisms, developed product-liability doctrine, and the ability to investigate technical evidence through formal civil procedure.

Weaknesses: Litigation costs, uncertainty over the application of different legal doctrines, and the possibility of inconsistent outcomes where several jurisdictions are involved.

5.4 Singapore model

Singapore is a major maritime and shipping centre whose regulatory approach combines maritime administration, commercial law, dispute resolution, and controlled technological development.

Its relevance lies in the ability to integrate operational safety with innovation and commercial certainty. Autonomous systems may require careful coordination among the maritime regulator, port authorities, ship operators, remote-control facilities, and technology suppliers.

For AI liability, the core civil-law questions remain contractual responsibility, negligence, causation, and available remedies. The regulatory question is whether the vessel and its remote-operation arrangements satisfy the applicable maritime safety requirements.

Strengths: A strong maritime-services ecosystem, commercial dispute-resolution infrastructure, and opportunities to coordinate innovation with operational supervision.

Weaknesses: National port-level regulation cannot by itself resolve every question arising from a collision in international waters or a defect supplied by a foreign manufacturer.

5.5 India's hybrid maritime model

India combines legislation with judicial precedent and international maritime obligations. Its framework includes the Merchant Shipping Act, 1958, as amended from time to time, the Admiralty (Jurisdiction and Settlement of Maritime Claims) Act, 2017, and applicable rules of civil procedure, tort law, contract law, and other legislation.

The precise statute applicable to a particular incident must be checked against the law in force at the time of the event.

Indian courts may be asked to determine maritime claims involving collisions, cargo damage, ship arrest, contractual disputes, and claims against foreign shipping interests. AI introduces additional issues concerning proof of software defects, identification of responsible parties, and the enforceability of contracts allocating technological risks.

Strengths: Statutory maritime jurisdiction, judicial remedies, and the ability to develop domestic rules alongside international maritime standards.

Weaknesses: AI-specific maritime liability questions may require clarification through legislation, regulations, and judicial interpretation. Cross-border access to software evidence and foreign defendants may also complicate proceedings.

6. Comparative table: how the models allocate AI maritime liability

The comparison shows that no single model completely eliminates AI maritime risk. The best framework combines international operational standards with clear national rules on liability, evidence, insurance, and compensation.

7. Eight important case laws relevant to civil liability and AI maritime regulation

Important qualification: The following decisions establish principles that may be relevant to AI-enabled maritime operations. They are not presented as judgments directly deciding liability for modern autonomous-ship AI failures. Their value lies in applying established rules on negligence, seaworthiness, causation, contractual risk, and maritime jurisdiction to new technological circumstances.

Case 1: M.V. Elisabeth v. Harwan Investment & Trading Pvt. Ltd.

Supreme Court of India · 1993 Supp (2) SCC 433

Facts and legal issue: The case concerned the jurisdiction of Indian courts over a foreign vessel and the scope of admiralty jurisdiction in India.

Decision and principle: The Supreme Court recognized the importance of admiralty jurisdiction as a body of law shaped by international maritime principles and emphasized the capacity of Indian courts to exercise maritime jurisdiction within the applicable legal framework.

Relevance to AI maritime regulation:

Suppose an autonomous foreign vessel enters an Indian port and is alleged to have caused damage through a defective navigation algorithm. The claimant may need to determine whether an Indian court can entertain the maritime claim and what remedies are available.

The case is relevant to questions concerning maritime jurisdiction, ship arrest where legally permissible, and access to remedies against foreign shipping interests.

Legal significance: It illustrates that technological innovation does not remove a vessel from established maritime jurisdictional principles. However, jurisdiction alone does not prove negligence or establish that the manufacturer or shipowner is liable.

Case 2: The CMA CGM Libra

UK Supreme Court · 20212021 UKSC 51

Facts: The container ship CMA CGM Libra grounded after following a passage that was unsafe. The relevant passage information had been prepared inadequately.

Legal issue: Whether the defective passage planning meant the vessel was unseaworthy, and whether the shipowner could rely on the statutory defence applicable to negligent navigation.

Decision and principle: The UK Supreme Court upheld the conclusion that the defective passage plan rendered the vessel unseaworthy. The obligation to exercise due diligence to make the ship seaworthy was not displaced simply because the defect concerned passage planning.

Relevance to AI:

Imagine that an AI navigation platform produces an unsafe route because its charts are outdated or its route-planning system has not been properly validated.

The case supports examining the ship's overall readiness for the intended voyage, including relevant navigational arrangements, rather than focusing exclusively on the moment the AI system made a wrong decision.

It also demonstrates why a shipowner cannot necessarily avoid responsibility merely by attributing a navigational failure to a particular operational process.

Legal significance: It is a particularly useful analogy for AI-enabled seaworthiness and the duty to exercise due diligence. Whether a specific AI defect makes a vessel unseaworthy would depend on the facts and applicable law.

Case 3: The Ocean Victory

UK Supreme Court · 20172017 UKSC 35

Facts: The vessel Ocean Victory was lost after encountering severe weather and conditions at Kashima port. The litigation concerned the allocation of responsibility for the loss under a bareboat charter and a related time charter.

Legal issue: Whether the relevant port conditions made the port unsafe and how contractual allocation of risk affected the parties' claims.

Decision and principle: The Supreme Court considered port safety in the context of the relevant charterparty provisions and the relationship between the contractual arrangements. The case illustrates that liability depends on the applicable contractual terms and the legally relevant characteristics of the risk.

Relevance to AI:

Suppose an autonomous ship is instructed to enter a port during adverse weather, although its certified operating envelope excludes the prevailing conditions. The incident may involve the shipowner, remote operator, port authority, and charterer.

The legal analysis would ask who had authority to approve the voyage, which party controlled the relevant decision, and whether the contractual allocation of operational risks applies.

Legal significance: The case helps explain why AI maritime liability cannot be allocated solely by asking who supplied the software. Contracts and the circumstances of the operation may be equally important.

Case 4: The Achilleas — Transfield Shipping Inc v Mercator Shipping Inc

UK House of Lords · 20082008 UKHL 48

Facts: A chartered vessel was redelivered late, affecting the shipowner's ability to perform a subsequent charter at the expected rate. The dispute concerned the extent of recoverable loss.

Legal issue: Whether the charterer was liable for the full claimed loss arising from the subsequent fixture or whether the damages were limited by the parties' assumed allocation of risk.

Decision and principle: The House of Lords addressed contractual remoteness and the significance of the risk the parties could reasonably be understood to have accepted. The decision is an important authority on the limits of contractual damages in commercial shipping.

Relevance to AI:

An AI-based cargo-scheduling system might delay a vessel and cause substantial losses under a chain of subsequent contracts. The fact that the AI malfunction caused a delay would not automatically mean that every downstream commercial loss is recoverable.

Courts may need to consider foreseeability, the applicable remoteness principles, contractual terms, and the allocation of risk between the parties.

Legal significance: The case is useful when evaluating claims for consequential commercial loss caused by AI-related delay. The result will depend on the particular contractual and legal framework.

Case 5: The Nicholas H — Marc Rich & Co AG v Bishop Rock Marine Co Ltd

UK House of Lords · 19951995 2 Lloyd's Rep 299

Facts: The litigation concerned a classification society's alleged negligence in connection with a vessel that subsequently suffered a casualty. Cargo interests sought to hold the classification society liable.

Legal issue: Whether the classification society owed the cargo interests a duty of care in negligence.

Decision and principle: The House of Lords rejected the negligence claim against the classification society, considering the relationship between the parties and relevant policy factors in determining whether the asserted duty of care should exist.

Relevance to AI:

AI maritime systems may be tested, assessed, or certified by independent organizations. If a vessel later suffers an AI-related casualty, injured cargo interests might attempt to sue the certifier or technical assessor.

This case illustrates that participation in a safety-related process does not automatically create a duty of care to every person who may suffer loss.

Legal significance: Liability of AI auditors, certification bodies, and independent assessors must be assessed under the applicable law. Their responsibilities cannot be assumed merely because their work concerns safety.

Case 6: Donoghue v Stevenson

House of Lords · 19321932 AC 562

Facts: A consumer allegedly suffered illness after consuming a drink containing a decomposed snail. The consumer had no direct contract with the manufacturer.

Legal issue: Whether a manufacturer could owe a duty of care to an ultimate consumer despite the absence of a direct contractual relationship.

Decision and principle: The House of Lords recognized the foundational modern negligence principle that a manufacturer may owe a duty of care to persons foreseeably affected by its products.

Relevance to AI:

A software developer may supply a collision-avoidance system through a manufacturer or integrator rather than directly to the shipowner. If the system is negligently designed and causes foreseeable harm, the absence of a direct contract with an injured third party may not necessarily defeat a negligence claim.

However, the existence of a duty and the availability of a remedy depend on the applicable jurisdiction's law, the relationship between the parties, and any relevant statutory regime.

Legal significance: The case helps explain the possible basis for direct negligence claims against AI developers and manufacturers even where the injured party did not purchase the software.

Case 7: The Wagon Mound (No. 1) — Overseas Tankship (UK) Ltd v Morts Dock & Engineering Co Ltd

Privy Council · 19611961 AC 388

Facts: Oil leaked into Sydney Harbour and spread across the water. Subsequent events led to a fire that damaged a wharf. The legal dispute concerned liability for the fire damage.

Legal issue: Whether the type of damage was sufficiently foreseeable to be recoverable in negligence.

Decision and principle: The Privy Council established the importance of reasonable foreseeability of the kind of damage when determining remoteness in negligence.

Relevance to AI:

Consider a cyberattack that compromises an autonomous vessel's navigation system and results in a collision, followed by an oil spill and environmental damage.

The court may have to examine which categories of loss were reasonably foreseeable, whether the relevant duty was breached, and whether the necessary causal connection exists.

The precise outcome would depend on the governing law and the applicable environmental and maritime liability conventions.

Legal significance: The case helps distinguish actual causation from legal responsibility for the extent and type of damage.

Case 8: Caparo Industries plc v Dickman

House of Lords · 19901990 2 AC 605

Facts: Investors relied on audited company accounts and alleged negligence by the auditors. The dispute concerned the scope of the duty of care owed to persons relying on the accounts.

Legal issue: Whether a duty of care existed in the circumstances and for the purpose for which the information was supplied.

Decision and principle: The House of Lords considered foreseeability, proximity, and whether it was fair, just, and reasonable to impose a duty of care.

Relevance to AI:

A company might market an AI navigation product as suitable for particular maritime conditions. An injured party could argue that it reasonably relied on the developer's safety representations.

The legal question would include whether the developer owed a duty to that claimant for the relevant purpose, rather than assuming that all foreseeable reliance gives rise to liability.

Legal significance: The case provides a framework for examining the scope of potential negligence liability for AI vendors, although the precise test varies by jurisdiction and has developed differently across legal contexts.

8. What the case laws collectively establish

Seaworthiness and due diligence

The CMA CGM Libra supports examining whether the ship and its navigational arrangements were fit for the intended voyage.

Contractual allocation of risk

The Ocean Victory and The Achilleas demonstrate the importance of contractual terms and the scope of recoverable loss.

Duties of manufacturers and assessors

Donoghue v Stevenson, Caparo, and The Nicholas H illustrate how duty of care depends on the parties, the circumstances, and the relevant legal test.

Causation and jurisdiction

The Wagon Mound addresses remoteness of damage, while M.V. Elisabeth illustrates the importance of maritime jurisdiction and effective remedies.

These authorities demonstrate that existing legal principles provide a substantial foundation for AI maritime liability. Nevertheless, conventional cases do not fully resolve questions such as whether an AI system can be treated as the legal equivalent of a master, how autonomous-system logs should be disclosed, or how liability should be allocated when the model changes through software updates.

Those issues require statutory clarification, regulatory standards, or future judicial decisions.

9. Civil liability models for AI maritime accidents

Three principal models can be considered when designing a civil liability regime for autonomous shipping.

9.1 Fault-based liability model

Under this model, the claimant generally must establish the elements required by the applicable law, such as duty, breach, causation, and legally recognized damage.

For example, a shipowner may be liable if it deploys an AI navigation system without adequate testing or fails to maintain a required safety system.

A manufacturer may be liable if a defect in its product causes foreseeable damage and the applicable legal requirements for liability are satisfied.

Advantages

Preserves the principle that liability should reflect legally relevant wrongdoing.

Encourages responsible development and operation.

Distinguishes unavoidable technical failures from negligent conduct.

Disadvantages

Technical opacity can make fault difficult to prove.

Multiple parties may blame one another.

Claimants may lack access to essential technical evidence.

9.2 Strict product liability model

Under a strict product-liability regime, a claimant may not need to prove the manufacturer's negligence. Instead, the claimant must establish the elements prescribed by the applicable law, typically including a qualifying product defect, damage, and causation.

This model can be relevant where defective AI software or hardware contributes to a maritime accident.

However, strict product liability is not a universal rule for every AI failure. Its availability depends on the governing jurisdiction, statutory scope, applicable defences, and the nature of the loss.

Advantages

Can improve access to compensation for victims.

Places appropriate responsibility on parties that design or supply defective products.

Encourages investment in product testing and quality control.

Disadvantages

Complex software may make the existence and timing of a defect difficult to establish.

Several suppliers may contribute to the same system.

Excessively broad liability may increase costs and discourage beneficial innovation.

9.3 Risk-based or enterprise-liability model

A risk-based model allocates duties according to operational control, capacity to prevent harm, and the risks associated with a system.

For example:

The shipowner would be responsible for safe deployment, maintenance, and appropriate insurance.

The operator would be responsible for compliance with operating procedures.

The developer would be responsible for applicable software-design and safety obligations.

The remote-control provider would be responsible for its assigned monitoring and intervention functions.

A future statute could combine fault-based liability with targeted presumptions or strict liability for specified risks.

This is a proposed regulatory model, not a universal rule already established in maritime law.

10. Proposed legal framework for AI maritime regulation

A workable system should combine international maritime standards with domestic civil liability law.

1. Mandatory risk assessment

Require documented assessment of collision risks, software failure, sensor errors, cyberattacks, communications loss, and foreseeable unsafe operating conditions.

2. Defined human responsibility

Identify the master, remote operator, shipowner, and system operator, including who has authority to intervene and what happens when communications fail.

3. Digital evidence and auditability

Preserve navigation decisions, sensor inputs, alerts, software versions, maintenance records, remote commands, and relevant cybersecurity logs.

4. Clear liability allocation

Establish how claims against shipowners, operators, manufacturers, and software suppliers interact, while preventing double recovery.

5. Insurance and compensation

Clarify insurance requirements, claims-handling arrangements, and any applicable liability limits or compulsory compensation mechanisms.

6. International harmonization

Align autonomous-vessel safety standards with maritime conventions and improve coordination of jurisdiction, evidence disclosure, and cross-border claims.

10.1 A practical liability test

When a maritime accident involves AI, a court or regulator could systematically examine the following questions:

Applicable law: Which national law, international convention, and contractual provisions govern the incident?

Relevant duty: What safety, contractual, statutory, or tort duty applied to each party?

Breach or defect: Did a party breach that duty, or did a qualifying product defect exist?

Causation: Did the breach or defect materially cause the accident or loss under the applicable legal test?

Defences and allocation: Do statutory defences, contractual exclusions, contributory fault, or other allocation rules apply?

Remedies: What compensation, limitation rules, insurance arrangements, or other remedies are available?

This sequence helps distinguish a technical malfunction from legally actionable misconduct. A system failure alone does not automatically prove that the manufacturer, shipowner, or remote operator is liable.

11. Special challenges in comparative regulation

A. Can AI itself be legally liable?

Under existing mainstream maritime liability frameworks, AI is not generally treated as an independent legal person that automatically bears civil liability for its decisions.

Responsibility is ordinarily attributed to legally recognized persons or entities under the applicable law. Creating separate legal personality for AI would be a major policy decision and is not a necessary precondition for regulating autonomous shipping.

B. How should courts handle the black-box problem?

An AI model may produce a decision that is difficult to reconstruct after an accident. A sound legal framework should require suitable recordkeeping and provide proportionate procedures for obtaining relevant evidence.

Courts may also need to address confidentiality, cybersecurity, trade secrets, and the protection of sensitive operational information.

C. What happens when several parties contribute to a collision?

Suppose a software developer provides a defective update, the shipowner installs it without required validation, and the remote operator ignores a warning.

Depending on the governing law, more than one party may be liable. The law must determine whether their responsibility is joint, several, apportioned, or otherwise allocated, as well as any rights of contribution between defendants.

D. How should international jurisdiction be determined?

A collision could involve a ship registered in one country, software developed in another, an operator based in a third, and damage in the waters of a fourth.

The applicable law and competent forum may depend on maritime jurisdiction, collision conventions, domestic legislation, contractual jurisdiction clauses, and conflict-of-laws rules. There is no universal answer that applies to every AI maritime accident.

12. Critical evaluation: which regulatory model is best?

No single jurisdiction's approach should be copied without adaptation.

The European model is valuable for structured product-safety requirements, codified liability rules, and risk-based oversight.

The English model provides mature maritime precedent and sophisticated reasoning about seaworthiness, negligence, contractual risk, and damages.

The United States model offers substantial experience in complex tort litigation and product-liability disputes.

Singapore's approach illustrates the importance of integrating maritime innovation, port operations, safety oversight, and commercial certainty.

India's hybrid model offers a statutory and judicial foundation that can be developed to address autonomous vessels and cross-border technological disputes.

The strongest solution is a hybrid international model with five essential features:

Common minimum safety and certification standards for autonomous ships.

Clearly allocated duties for shipowners, masters, remote operators, and AI suppliers.

Effective evidence-preservation and disclosure requirements.

Accessible compensation supported by appropriate insurance arrangements.

Harmonized rules for cross-border jurisdiction and liability, without unnecessarily restricting technological innovation.

This model would reduce the risk that the parties responsible for a maritime accident escape accountability simply because the immediate decision was made by software.

13. Conclusion

Civil law and comparative AI maritime regulation intersect at the fundamental question of how responsibility should be allocated when increasingly autonomous systems cause harm.

Traditional principles of negligence, seaworthiness, product liability, contractual responsibility, causation, and damages remain relevant. The cases discussed—including M.V. Elisabeth, The CMA CGM Libra, The Ocean Victory, The Achilleas, The Nicholas H, Donoghue v Stevenson, The Wagon Mound, and Caparo—provide important foundations for analyzing maritime AI disputes, although none should be mistaken for a direct judicial determination of liability under a comprehensive autonomous-ship AI regime.

The international framework has also advanced: the IMO adopted its non-mandatory MASS Code in 2026, creating a significant step toward globally coordinated autonomous-ship safety regulation. Civil liability, however, still requires careful integration with the applicable national laws, international conventions, and contractual arrangements.

Ultimately, the objective should be to regulate AI maritime systems according to their risks while ensuring that human and corporate responsibility remains identifiable, injured parties have meaningful remedies, and international shipping retains consistent safety standards.

Suggested legal research structure

For an academic paper, dissertation, or seminar, the topic can be organized under these headings:

Introduction and conceptual framework.

Civil liability principles applicable to AI maritime systems.

International maritime conventions and the IMO MASS Code.

Comparative analysis of European, English, American, Singaporean, and Indian models.

Judicial precedents and their application to autonomous shipping.

Liability, causation, insurance, and digital evidence.

Regulatory gaps and proposals for reform.

Findings and conclusion.

The central research question is: How can comparative civil liability systems allocate responsibility fairly and predictably for damage caused by AI-enabled maritime operations while maintaining international maritime safety and technological innovation?