Civil Law and Comparative AI Intellectual Property Models

1. Introduction

Artificial Intelligence (AI) has created significant challenges for intellectual property (IP) law, particularly in determining who owns AI-generated works, whether AI-generated inventions are patentable, how copyrighted material may be used to train AI systems, and who is legally responsible when AI produces infringing content.

Traditionally, intellectual property law has been built around human creativity and inventorship. AI systems, however, can generate literary works, paintings, music, software code, technical designs, and potentially inventive solutions with varying degrees of human involvement. This development raises a fundamental question:

Can an AI system create intellectual property, and if so, who should own the resulting rights?

The answer differs across legal systems. Civil law jurisdictions, such as Germany, France, and Japan, generally operate within statutory frameworks that emphasize human authorship, personal rights, and codified rules. Common law jurisdictions, such as the United States and the United Kingdom, rely on a combination of legislation, judicial precedent, and statutory interpretation, although their approaches to AI-generated works differ substantially.

A comparative study must distinguish between four separate questions:

Authorship: Who is legally recognized as the author of an AI-generated work?

Ownership: Who owns the economic rights in the output?

Patentability: Can an AI system be recognized as an inventor?

Training and infringement: Can copyrighted works be used to train AI models, and who is liable if an output infringes existing rights?

These questions are related but do not necessarily have the same answer. A person may own the copyright in a work created with AI assistance without owning every element generated by the AI. Similarly, a company may own contractual rights in AI output without that output qualifying for copyright protection.

2. Meaning of civil law and comparative AI intellectual property models

2.1 Civil law model

Civil law refers to a legal tradition substantially based on codified statutes and systematic legal principles, historically influenced by Roman law. It is found in countries such as France, Germany, Japan, and many other jurisdictions.

In intellectual property disputes involving AI, civil law systems commonly emphasize:

Statutory definitions of authorship and inventorship.

The connection between authorship and a natural person.

Moral rights, including attribution and protection against certain distortions of a work.

Codified rules governing copyright, patents, and ownership.

Judicial interpretation of legislation within an established statutory framework.

However, civil law countries are not identical. French copyright law places particular emphasis on the author's personality, while German law contains an express rule identifying the creator as the author. Japan has its own statutory framework and government policies concerning generative AI.

2.2 Common law model

Common law systems traditionally give an important role to judicial precedent, alongside legislation. The United States and the United Kingdom are important examples.

Their AI intellectual property models differ:

The United States generally requires human authorship for copyright protection and a natural person to qualify as a patent inventor.

The United Kingdom has a special statutory provision for certain computer-generated works, identifying the person who made the necessary arrangements for their creation as the author.

Both systems recognize that AI-assisted works may receive protection when the relevant legal requirements are satisfied.

2.3 Comparative models at a glance

The central distinction is not that civil law always prohibits AI-related protection while common law always permits it. Rather, each jurisdiction determines how its existing legal concepts apply to AI-generated material.

3. The principal AI intellectual property models

3.1 Human-authorship model

Under this model, copyright protection depends on a human being making the legally required creative contribution. AI is treated as a tool rather than an independent author.

For example, a graphic designer uses an AI image generator to produce ten possible illustrations, selects one, changes its composition, redraws significant elements, and incorporates the result into a publication.

The designer may obtain copyright protection in the human-created elements if the applicable originality requirements are met. The AI-generated elements do not automatically become copyrightable merely because the designer selected or downloaded them.

Advantages

Preserves the traditional connection between creativity and legal authorship.

Provides a basis for determining who can exercise moral rights.

Avoids granting rights to entities that cannot independently hold or exercise them under existing law.

Limitations

Determining the extent of human creative contribution can be difficult.

The model may leave fully autonomous AI output without copyright protection.

Different courts may draw different boundaries between human creativity and machine-generated material.

3.2 Computer-generated-work model

This model permits a statutory attribution of authorship to a person who made the necessary arrangements for creating a computer-generated work, even when the work does not have a conventional human author.

The United Kingdom provides a significant example through section 9(3) of the Copyright, Designs and Patents Act 1988. It defines the author of a computer-generated literary, dramatic, musical, or artistic work as the person who undertook the arrangements necessary for its creation.

This does not mean that every AI output automatically qualifies for copyright. The statutory requirements, including the meaning of a computer-generated work and applicable originality rules, must still be considered.

The model offers a possible solution to the ownership problem, but questions remain about which person made the necessary arrangements: the programmer, the system operator, the employer, the commissioning party, or another person.

3.3 Investment-based or incentive model

Under this approach, intellectual property law is considered partly in terms of the investment needed to develop and operate AI systems.

A developer may invest in computing infrastructure, training data, research, model development, and deployment. A user may invest in selecting prompts, refining outputs, editing content, and preparing a finished work.

The central policy question is whether legal exclusivity should reward such investment even when the resulting output lacks sufficient human creativity.

Investment alone does not ordinarily establish copyright authorship. However, it may support protection through contracts, trade secrets, database rights where applicable, or other legally recognized rights.

3.4 Hybrid model

The hybrid model distinguishes among human-created material, AI-generated material, and the human contribution to an AI-assisted final product.

For example, a novelist might use AI to suggest alternative plot structures, generate preliminary descriptions, and identify inconsistencies. The novelist then writes the final prose, creates the characters, restructures the narrative, and edits the manuscript.

The resulting work may receive copyright protection for the novelist's original expression, while purely machine-generated portions may receive different treatment.

This model is particularly practical because it does not require treating AI as either completely irrelevant or equivalent to a human author.

4. Comparative analysis of major jurisdictions

4.1 India

India is especially relevant because its copyright and patent statutes were drafted before modern generative AI became widespread.

A. Copyright law

The Copyright Act, 1957, contains several provisions relevant to AI-generated works:

Section 2(d)(vi): Defines the author of a computer-generated literary, dramatic, musical, or artistic work as the person who causes the work to be created.

Section 13: Establishes the categories of works in which copyright may subsist.

Section 14: Defines the exclusive rights associated with copyright.

Section 17: Contains rules concerning first ownership of copyright and specified exceptions.

Section 57: Recognizes moral rights, including rights of authorship and protection against certain prejudicial distortions or modifications.

Section 2(d)(vi) makes India distinctive because it expressly addresses computer-generated works. Nevertheless, the application of this provision to modern generative AI raises unresolved questions.

For example, if a user enters a short prompt into a large language model, is the user the person who causes the work to be created? Could the relevant person instead be the developer or operator of the system? Does the answer depend on the user's creative control over the final output?

These issues require interpretation of the statute and cannot be resolved merely by pointing to the provision's existence.

B. Patent law

Under the Patents Act, 1970, inventorship and entitlement to apply for a patent are governed principally by sections 2(1)(y), 6, and 7, together with other applicable provisions.

Section 6 identifies persons entitled to apply for a patent, including the true and first inventor and certain persons deriving title from that inventor. Indian law has not established a general rule granting patent inventorship to an AI system.

An AI-assisted invention may still be patentable if the statutory requirements are met and a legally recognized inventor and entitled applicant can be identified.

C. Practical position

India has a statutory basis for addressing computer-generated copyright works, but this should not be interpreted as an automatic grant of copyright to every output produced by generative AI. The extent of human involvement, statutory interpretation, originality, first ownership, and contractual arrangements remain important.

4.2 France

French copyright law places substantial emphasis on the human author's personality and creative expression.

The Intellectual Property Code, particularly Article L.111-1, recognizes the author's intellectual property right arising from the creation of a work. French law also gives considerable importance to moral rights.

Under this approach, the main challenge is establishing that a work reflects the personal intellectual contribution required for copyright protection.

A fully autonomous AI output therefore raises a more difficult authorship problem than a work in which a human author makes significant expressive decisions.

For instance, a person who merely types a generic prompt may have difficulty establishing that the final output reflects their own original creative expression. By contrast, substantial human selection, arrangement, modification, and original composition may support protection of the resulting human-created elements.

France also operates within the European Union's copyright framework, including rules relevant to text and data mining. Those rules must be considered separately from whether an AI output itself qualifies for copyright.

4.3 Germany

German copyright law offers a particularly clear example of the human-centered approach.

Section 7 of the German Copyright Act identifies the author as the creator of the work. German law also protects the author's relationship with the work through moral and economic rights.

This approach makes it difficult to treat an autonomous AI system as an author under the ordinary copyright framework.

However, AI assistance does not necessarily prevent protection. A human creator's original contribution may qualify where the person exercises sufficient creative influence over the final expression.

German law also distinguishes copyright from other legal mechanisms. Contractual rights, technical protection measures, trade secrets, and certain other forms of protection may be relevant even where copyright does not subsist in an entirely machine-generated output.

4.4 United States

The United States follows a human-authorship approach to copyright, but it has also developed detailed administrative guidance on works containing AI-generated material.

The Copyright Act, 1976, protects original works of authorship fixed in a tangible medium of expression. Under the prevailing interpretation, a human being must be the author.

The US Copyright Office distinguishes between:

Works produced entirely by AI without sufficient human creative contribution.

Works in which a human selects or arranges AI-generated material creatively.

Works in which AI assists a human author who independently contributes original expression.

Only the qualifying human-authored aspects are eligible for copyright protection under this approach.

For patents, the Supreme Court's 2022 decision in Thaler v. Vidal confirmed that the statutory concept of an inventor under the Patent Act refers to a natural person. The distinction between copyright authorship and patent inventorship is important: both presently require human legal participation, but they arise under separate statutes and tests.

4.5 United Kingdom

The United Kingdom combines a human-centered approach to patent inventorship with a special copyright provision for computer-generated works.

Under section 9(3) of the Copyright, Designs and Patents Act 1988, the person who undertakes the necessary arrangements for creating a computer-generated work may be treated as its author.

This is a statutory attribution model rather than a declaration that AI has legal personality.

Its advantages include potential certainty in commercial situations where a work is generated without conventional human authorship. Its weaknesses include uncertainty about who undertook the necessary arrangements and how the provision applies to increasingly autonomous generative systems.

The UK Supreme Court's decision in Thaler v. Comptroller-General confirms that this special copyright rule cannot simply be transferred to patent law. The patent statute requires a human inventor.

4.6 Japan and the European Union

Japan has a statutory copyright framework and has addressed generative AI through policy guidance, including discussion of copyright limitations for AI training. Whether a use is permitted depends on the relevant provisions and circumstances; it should not be assumed that all AI training is automatically lawful.

The European Union provides a further comparative model through its harmonized copyright rules, including the Digital Single Market Directive's text-and-data-mining provisions in Articles 3 and 4.

Broadly, Article 3 concerns qualifying scientific research by specified research organizations and cultural heritage institutions. Article 4 provides a broader text-and-data-mining exception subject to conditions, including the ability of rightsholders to reserve rights in the manner provided by law.

These rules primarily address the use of material for mining. They do not, by themselves, establish copyright ownership in an AI-generated output.

The EU's AI Act also creates transparency and copyright-policy obligations relevant to general-purpose AI providers. Such obligations should be distinguished from the separate question of whether a particular output is protected by copyright.

5. Important case laws on AI and intellectual property

The following cases provide the principal legal foundations for understanding AI authorship, inventorship, originality, and copyright infringement. Some directly concern AI, while others establish earlier principles that courts apply when considering AI-related disputes.

Case 1. Thaler v. Perlmutter — United States, 2025

United States Court of Appeals for the District of Columbia Circuit · 2025

Facts: Stephen Thaler sought copyright registration for an image titled A Recent Entrance to Paradise, which he claimed had been generated autonomously by an AI system called the Creativity Machine. He identified the AI system as the author.

Legal issue: Can an AI system qualify as an author under US copyright legislation?

Decision: The Court of Appeals affirmed the refusal of copyright registration because the Copyright Act requires human authorship. The decision was delivered on 18 March 2025.

Legal principle: An AI system cannot independently qualify as the author of a copyright-protected work under the applicable US law.

Relevance to comparative law: The decision illustrates the US human-authorship model. It contrasts with the UK's statutory provision for certain computer-generated works and India's express inclusion of a person who causes a computer-generated work to be created.

The decision does not mean that all works produced using AI are unprotected. Human contributions may qualify for copyright where they satisfy the applicable requirements.

Case 2. Thaler v. Comptroller-General of Patents, Designs and Trade Marks — United Kingdom, 2023

UK Supreme Court · 20232023 UKSC 49

Facts: Thaler filed patent applications identifying DABUS, an AI system, as the inventor. He argued that, as the system's owner, he was entitled to obtain patents for its inventions.

Legal issue: Can an AI machine be named as an inventor under the Patents Act 1977, and can ownership of that machine establish entitlement to the resulting patents?

Decision: The UK Supreme Court unanimously dismissed the appeal. It held that an inventor under the relevant legislation must be a natural person. Ownership of the AI system did not, by itself, entitle Thaler to the patents.

Legal principle: Patent inventorship and ownership are separate questions. Owning the machine that allegedly generated an invention does not automatically establish ownership of the invention or the right to obtain a patent.

Relevance: This is a leading case demonstrating that AI patent inventorship remains distinct from the question of whether an AI-assisted invention can be patented when a human inventor is properly identified.

Case 3. Thaler v. Vidal — United States, 2022

United States Court of Appeals for the Federal Circuit · 43 F.4th 1207

Facts: Thaler sought patent protection for inventions allegedly generated by DABUS without a human inventor. The US Patent and Trademark Office rejected the applications because DABUS was identified as the inventor.

Legal issue: Does the US Patent Act permit an AI system to qualify as an inventor?

Decision: The Federal Circuit held that the statutory term “individual” in the Patent Act refers to a natural person. An AI system therefore cannot be named as an inventor under the relevant provisions.

Legal principle: Patent inventorship requires a natural person under the US statute.

Relevance: Read together with the UK Supreme Court's decision in Thaler, this case shows convergence between two jurisdictions that otherwise have different legal traditions.

The decisions concern the legal identity of an inventor, not a blanket prohibition on patents for inventions developed with AI assistance.

Case 4. Eastern Book Company v. D.B. Modak — India, 2008

Supreme Court of India · (2008) 1 SCC 1

Facts: The dispute concerned copyright claims over edited and published Supreme Court judgments. The publisher argued that its editorial work justified copyright protection in the resulting text.

Legal issue: What degree of originality is necessary for copyright protection in India?

Decision: The Supreme Court rejected the idea that mere labour, expense, or skill automatically creates copyright. The work must involve a sufficient degree of original intellectual effort and creativity.

Legal principle: Indian copyright law requires originality beyond mere mechanical effort or the application of skill and labour.

Relevance to AI: This is not an AI case, but its reasoning is important when evaluating AI-generated content. The fact that a person spends money on an AI system, writes prompts, or performs routine editing does not automatically establish copyright in the output.

A human contribution involving original selection, arrangement, expression, or substantial creative modification may be more persuasive.

The judgment also cautions against assuming that a computer-generated work qualifies for copyright merely because resources were invested in producing it.

Case 5. Burrow-Giles Lithographic Co. v. Sarony — United States, 1884

US Supreme Court · 111 U.S. 53

Facts: A lithographic company reproduced a photograph of Oscar Wilde. Photographer Napoleon Sarony claimed copyright in the photograph.

Legal issue: Could a photograph qualify as an original work of authorship when a camera was used to produce it?

Decision: The Supreme Court upheld copyright protection, emphasizing Sarony's creative involvement in posing the subject, arranging the scene, selecting lighting, and determining the composition.

Legal principle: A technological instrument can be used to create a copyright-protected work where the human creator contributes original intellectual conception and expression.

Relevance to AI: The analogy is particularly useful because both photography and generative AI involve technology that contributes to the production of an image.

However, the analogy has limits. In Sarony, the photographer's creative decisions were established. In autonomous AI generation, the extent to which the human user determines the final expressive details may be much harder to establish.

Case 6. Feist Publications, Inc. v. Rural Telephone Service Co. — United States, 1991

US Supreme Court · 499 U.S. 340

Facts: Feist copied factual listings from a telephone directory published by Rural Telephone Service. Rural claimed copyright in the directory.

Legal issue: Is the expenditure of effort sufficient to establish copyright in a compilation of facts?

Decision: The Supreme Court held that the listings lacked the necessary originality for the claimed protection. Copyright does not reward labour alone; independent creation and at least a minimal degree of creativity are required.

Legal principle: Originality requires independent creation and a minimal creative contribution, not simply substantial investment.

Relevance to AI: The case helps distinguish between generating a large volume of content and creating original expression. An AI developer's investment in infrastructure or a user's expenditure on prompts does not automatically confer copyright in every output.

It also shows why copyright protection for databases, training materials, and AI-generated content must be examined separately.

Case 7. Naruto v. Slater — United States, 2018

US Court of Appeals for the Ninth Circuit · 888 F.3d 418

Facts: A macaque monkey took photographs using a photographer's unattended camera. The animal's representatives subsequently brought copyright infringement proceedings on its behalf.

Legal issue: Could a non-human animal bring a copyright infringement claim under the US Copyright Act?

Decision: The Ninth Circuit held that animals could not sue under the Copyright Act because the statute did not expressly authorize them to do so.

Legal principle: Creating something does not automatically establish that the creator has the legal status or statutory rights required to own and enforce copyright.

Relevance to AI: The case provides an analogy for the distinction between factual creation and legal authorship. Even if a machine produces an image, it does not follow that the machine can own copyright or enforce legal rights.

The decision does not directly decide AI authorship, but it illustrates the importance of statutory recognition and legal capacity.

Case 8. Aalmuhammed v. Lee — United States, 2000

US Court of Appeals for the Ninth Circuit · 202 F.3d 1227

Facts: A contributor to the film Malcolm X claimed joint authorship based on his substantial creative contributions, including script revisions and other work on the production.

Legal issue: When does a person's contribution to a collaborative work establish joint authorship?

Decision: The court rejected the joint-authorship claim, examining factors including control over the work, objective manifestations of shared intent, and the nature of the creative contribution.

Legal principle: Contributing ideas, suggestions, or assistance does not automatically establish joint authorship.

Relevance to AI: A person who contributes prompts, reviews outputs, or gives general instructions may not necessarily qualify as an author of the resulting work. The legal significance of the person's contribution depends on the applicable jurisdiction and the nature of the creative control exercised.

This case is useful for distinguishing between operating an AI tool and making a legally sufficient creative contribution.

Case 9. Infopaq International A/S v. Danske Dagblades Forening — European Union, 2009

Court of Justice of the European Union · Case C-5/08

Facts: Infopaq operated a system that scanned newspapers and extracted short text extracts for information-monitoring purposes.

Legal issue: Could the reproduction of short extracts of text constitute copyright infringement?

Decision: The Court held that even a short extract can qualify as a reproduction of part of a protected work if it contains elements expressing the author's own intellectual creation.

Legal principle: Copyright analysis depends on the original expression reproduced, not merely on the length of the extract.

Relevance to AI: AI-generated text can raise infringement questions even when it is relatively short. A short passage may reproduce protected expression, while a longer passage may consist primarily of unprotected ideas or factual material.

The case is relevant to both AI training and AI output, although it does not itself determine the legality of training generative AI models.

Case 10. Cofemel — Sociedade de Vestuário SA v. G-Star Raw CV — European Union, 2019

Court of Justice of the European Union · Case C-683/17

Facts: The dispute concerned whether clothing designs could qualify as copyright-protected works in addition to receiving protection under other intellectual property regimes.

Legal issue: What conditions must a subject matter satisfy to qualify as a copyright-protected work under EU law?

Decision: The Court emphasized that copyright protection requires original subject matter that reflects the author's own intellectual creation and is expressed in a manner identifiable with sufficient precision and objectivity.

Legal principle: A work must satisfy the applicable originality and identifiability requirements; aesthetic merit is not an independent requirement for copyright protection.

Relevance to AI: The decision helps explain why a generated image or design must be assessed for originality and legally recognizable expression. It does not, however, resolve whether a fully autonomous AI system can qualify as an author.

Case-law summary

6. Copyright infringement and AI training

AI intellectual property law is not limited to deciding who owns generated content. A separate and increasingly important issue concerns the use of existing copyrighted works to train AI systems.

6.1 The training process

Generative AI models may be trained on large collections of text, photographs, artwork, music, software, and other material.

During training, a system may process or reproduce data in ways that engage copyright law. Whether those acts infringe copyright depends on the jurisdiction, the particular technical process, the nature of the source material, applicable exceptions, licensing, and other relevant facts.

Three separate legal questions should be examined:

Input legality: Was the training material lawfully obtained and used?

Training legality: Did the relevant copying or computational analysis fall within a licence or statutory exception?

Output legality: Does the generated material reproduce protected expression or otherwise infringe an existing right?

A lawful training process does not necessarily make every output lawful. Equally, an output that does not infringe copyright does not automatically establish that the training process was lawful.

6.2 The comparative position

India: Section 52 of the Copyright Act, 1957, specifies acts that do not constitute infringement in defined circumstances. Its application to commercial generative AI training is not automatically settled by the existence of these exceptions. A detailed analysis of the relevant subsection, the acts performed, and the circumstances is necessary.

European Union: The Digital Single Market Directive provides separate text-and-data-mining exceptions, including the scientific-research exception and the broader exception subject to conditions such as rights reservation. The legal basis must be assessed for the particular use.

United States: Fair use may be relevant to certain training activities, but it is a fact-specific doctrine. The purpose and character of the use, the nature of the copyrighted work, the amount used, and the effect on the potential market are among the statutory considerations. No general rule makes all AI training either lawful or unlawful.

United Kingdom: Copyright exceptions and the specific statutory provision for computer-generated works must be analyzed independently. The existence of section 9(3) does not create a blanket authorization to train AI on copyrighted material.

France and Germany: EU law is particularly important because national rules operate alongside the relevant EU copyright framework. The availability of a text-and-data-mining exception does not itself confer copyright on generated outputs.

6.3 What if AI reproduces an existing work?

Suppose an AI image generator produces an image closely resembling a copyrighted illustration.

A legal assessment would ask:

Is the original illustration protected by copyright?

Does the generated image reproduce a substantial part of protected expression?

Is the similarity attributable to unprotected ideas, style, facts, or conventions, or to protected expressive elements?

Does any licence, exception, or other defence apply?

Which party is legally responsible under the applicable law?

The fact that the AI produced the image does not, by itself, eliminate potential infringement.

7. AI patents: inventorship, ownership, and patentability

Patent law presents a distinct set of challenges.

Copyright protects qualifying original expression. Patents protect qualifying inventions that satisfy statutory requirements, such as novelty, inventive step or non-obviousness, and industrial applicability or utility, depending on the jurisdiction.

Three concepts must be distinguished.

Inventorship identifies the person who legally qualifies as the inventor.

Ownership determines who holds the relevant proprietary rights, potentially through assignment, employment rules, or other legal mechanisms.

Patentability determines whether the invention meets the substantive and procedural requirements for patent protection.

An AI system might be instrumental in discovering a new pharmaceutical compound or designing a manufacturing process. Nevertheless, a patent application must still satisfy the applicable inventorship requirements and identify a legally recognized inventor.

The DABUS decisions in the United States and United Kingdom establish that an AI system cannot be designated as the inventor under the statutes considered in those cases.

They do not establish that every invention developed using AI is unpatentable. If a human makes a legally sufficient inventive contribution, the resulting invention may qualify for patent protection, subject to all other requirements.

8. Civil liability, ownership disputes, and contractual protection

Intellectual property ownership is only one part of the legal analysis. AI-generated works can also produce disputes involving contracts, confidentiality, licensing, and responsibility for harm.

8.1 Contractual ownership

AI service providers may include contractual terms dealing with ownership or permitted use of generated content. These terms can allocate rights between a provider and a user, but they cannot necessarily create copyright where the law recognizes none.

For example, a contract may permit a customer to commercially exploit an AI-generated logo. That permission does not automatically guarantee that the logo qualifies for copyright or that it does not infringe another person's rights.

8.2 Employment and commissioned works

Where an employee or contractor uses AI in creating a work, ownership may depend on statutory rules, employment arrangements, assignments, and the nature of the contribution.

A company should not assume that paying for an AI-generated work automatically gives it every intellectual property right associated with that work.

8.3 Confidential information and trade secrets

Even when copyright protection is unavailable, confidential prompts, proprietary datasets, model weights, source code, and technical processes may qualify for protection under applicable confidentiality or trade-secret rules.

Such protection generally depends on the relevant legal requirements, including confidentiality measures and the nature of the information.

8.4 Liability for infringing outputs

Depending on the jurisdiction and facts, potential claims may involve direct infringement, authorization, contractual liability, or other recognized legal grounds. The liability of the model developer, deployer, user, or commercial distributor cannot be determined solely by identifying who entered the prompt.

9. Critical comparative evaluation

The principal legal models can be evaluated according to five criteria.

Which model is preferable?

The human-authorship model provides a strong connection between legal protection and human creative activity. Its principal weakness is the possibility that valuable autonomous AI output may remain outside copyright protection.

The computer-generated-work model can improve commercial certainty, but its effectiveness depends on clear rules for identifying the person responsible for the necessary arrangements.

The hybrid model offers a practical middle ground. It recognizes that human creators increasingly use AI as part of their workflow without assuming that every output is entirely human-authored or entirely machine-generated.

However, no single model resolves every issue. Copyright authorship, patent inventorship, AI training, moral rights, and liability require distinct legal analyses.

10. Practical examples

Example A: AI-generated painting

An individual enters a simple prompt and downloads the resulting image without further changes.

Under the US human-authorship model, the image may lack copyright protection if it contains no sufficient human-authored expression.

Under the UK's computer-generated-work provision, the statutory attribution question must be considered, including whether the necessary arrangements were made by the person claiming authorship.

Under Indian law, section 2(d)(vi) is relevant, but its application to modern generative AI and the originality requirements must be considered.

Under French and German law, the absence of sufficient human creative contribution may make copyright protection difficult to establish.

Example B: AI-assisted novel

An author uses AI to brainstorm characters, identify plot inconsistencies, and suggest alternative scenes, then writes and revises the manuscript.

The author's original written expression may qualify for copyright protection. The mere use of AI does not disqualify the entire manuscript. The status of any material generated by AI and incorporated without meaningful human authorship must be assessed separately where the relevant law requires it.

Example C: AI-designed invention

A scientist uses an AI system to identify a promising chemical compound and then contributes to developing and specifying the invention.

The scientist's role must be evaluated under the relevant patent law to determine whether the person qualifies as an inventor. The AI system cannot simply be named as the inventor in jurisdictions applying the rules established in the DABUS decisions.

Example D: AI-trained commercial product

A business trains a model on copyrighted images and sells generated artwork.

The business must consider the legality of its training practices, the applicable licences and exceptions, and whether the outputs infringe existing rights. Ownership or licensing terms between the business and its customers do not independently resolve those questions.

11. Recommendations for legal reform

A balanced legal framework should consider the following reforms.

1. Clarify AI authorship: Legislatures should specify how much human creative contribution is required and whether any special attribution rule is appropriate for fully computer-generated works.

2. Distinguish output ownership from copyright: Contracts may allocate commercial use and other interests, but should not misleadingly suggest that copyright necessarily exists in every AI output.

3. Improve training-data transparency: Rules should provide appropriate clarity about the use of protected works for training, the availability of licences, and applicable rights-reservation mechanisms.

4. Preserve human inventorship rules or reform them expressly: If a jurisdiction wishes to recognize AI inventorship, it should do so through clear legislation rather than leaving the matter to uncertain interpretation of existing statutes.

5. Recognize moral rights appropriately: Reforms should consider attribution, integrity, and the practical difficulties of applying personality-based rights to autonomous machine output.

6. Encourage documentation: Creators and businesses should preserve records of prompts, drafts, edits, source material, licences, and human decision-making. These records can help establish the nature of the contribution and the basis for ownership claims.

7. Promote international consistency: Since AI systems operate across borders, greater coordination would reduce uncertainty concerning licensing, infringement, authorship, and enforcement.

12. Conclusion

Civil law and comparative AI intellectual property models reveal a common legal challenge: traditional IP systems were largely designed around human authors and inventors, whereas generative AI can produce content and technical solutions with varying degrees of human involvement.

The case law demonstrates several important principles. Thaler v. Perlmutter confirms the US human-authorship requirement; Thaler v. Comptroller-General and Thaler v. Vidal establish that AI cannot qualify as an inventor under the relevant UK and US patent statutes; Eastern Book Company v. D.B. Modak emphasizes originality under Indian copyright law; and Burrow-Giles, Feist, Naruto, Aalmuhammed, Infopaq, and Cofemel provide further guidance on creativity, legal authorship, originality, and protected expression.

The most important distinction is between the ability of AI to generate material and the legal requirements for protecting that material. A work may be commercially valuable without qualifying for copyright, and an invention may be developed with AI without making the AI a legally recognized inventor.

A carefully designed hybrid framework—one that protects genuine human creativity, clarifies the treatment of autonomous output, respects existing rights, and provides transparent rules for training and commercial use—offers a promising direction for future reform.

Academic note: This is a comparative legal overview, not jurisdiction-specific legal advice. The cases discussed establish principles relevant to AI intellectual property, but not all directly concern generative AI. For legal research or litigation, the current legislation and subsequent judicial developments in the relevant jurisdiction should be checked.