Attribution for AI assisted outputs

Introduction

Attribution for AI-assisted outputs refers to the legal, ethical and academic principles governing the identification of human and artificial intelligence contributions in works created with the assistance of AI technologies. These outputs may include written content, research papers, photographs, paintings, software code, scientific publications, musical compositions and other creative or intellectual works.

The rapid development of generative AI systems has created important questions concerning authorship, ownership, originality, accountability and intellectual property rights. Traditional copyright laws generally recognise human creators as authors, whereas AI systems can independently generate content based on user instructions, training data and computational processes.

Attribution becomes particularly important when AI-generated material is used in academic research, commercial publications, professional services and scientific investigations. Failure to disclose AI assistance may create concerns involving plagiarism, academic misconduct, copyright infringement and misleading representations of human authorship.

In India, attribution for AI-assisted outputs is primarily examined under the Copyright Act, 1957, the Information Technology Act, 2000, contractual principles and applicable institutional policies. Internationally, the United States Copyright Office, European Union institutions and international intellectual property organisations have developed important approaches to AI-generated works.

Legal framework governing attribution for AI-assisted outputs

1. Copyright Act, 1957 (India)

The Copyright Act, 1957, provides the principal legal framework governing authorship, ownership and protection of creative works in India.

Section 2(d) defines the term author for different categories of copyrightable works.

Importantly, Section 2(d)(vi) provides that, in relation to a literary, dramatic, musical or artistic work which is computer-generated, the author means the person who causes the work to be created.

This provision is particularly relevant to AI-assisted outputs because it recognises computer-generated works within the statutory definition of authorship.

However, the application of this provision to modern generative AI systems remains legally unsettled. Questions may arise concerning whether the relevant human contribution is sufficient to establish authorship and who should be recognised when several people contribute to the creation of an AI-assisted work.

Section 13 establishes the categories of works in which copyright subsists, subject to statutory requirements.

Section 14 defines the exclusive rights associated with copyright ownership.

Therefore, attribution and copyright ownership are related but separate questions. A person may deserve acknowledgement for contributing to an AI-assisted work without necessarily owning all copyright interests in that work.

2. Copyright and related rights under international law

The Berne Convention for the Protection of Literary and Artistic Works, 1886, establishes international standards for copyright protection.

It recognises the importance of authors' rights and provides protection for literary and artistic works.

Article 6bis recognises moral rights, including the right to claim authorship and object to certain prejudicial modifications of a work.

However, the Convention was developed around human creative activity and does not establish a comprehensive international framework for AI-generated authorship.

The Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS), 1994, also establishes minimum standards for copyright protection among WTO members.

Neither instrument provides a universally accepted solution concerning the attribution of outputs generated by autonomous AI systems.

Human authorship and AI-generated content

One of the most important legal questions is whether AI-generated content can receive copyright protection without meaningful human creative contribution.

Traditional copyright systems generally require a legally recognisable author and protect original expression rather than mere ideas, facts or automated processes.

In AI-assisted creation, different levels of human involvement may exist.

1. Fully AI-generated outputs

Fully AI-generated outputs are produced primarily through automated processes without meaningful human creative control over the final expression.

Examples include automatically generated images, articles or musical compositions.

In such circumstances, establishing human authorship and copyright protection may be difficult, depending on the applicable national law.

2. AI-assisted human-created outputs

AI-assisted works involve substantial human creative participation.

A person may provide original ideas, select materials, modify generated content, arrange different elements and make creative decisions concerning the final work.

Such contributions may support copyright protection where the applicable legal requirements are satisfied.

3. Human editing and modification

Human editing may create an independently protectable contribution.

For example, a researcher may use AI to prepare an initial draft and subsequently introduce original analysis, reorganise arguments and develop original conclusions.

Copyright protection may extend to the human-authored elements, even if the underlying AI-generated material does not independently qualify for protection.

Attribution requirements in academic and scientific research

Academic institutions and research organisations increasingly require disclosure of AI assistance.

Attribution is important because academic research depends on transparency, originality and accountability.

Researchers must distinguish between AI assistance and genuine human intellectual contributions.

Common disclosure practices include:

Identifying the AI tool used.

Explaining the purpose for which it was used.

Disclosing significant AI-generated material.

Verifying factual claims and references.

Identifying human authors responsible for the final publication.

Ensuring that AI systems are not falsely presented as accountable human authors.

The International Committee of Medical Journal Editors (ICMJE) states that AI tools should not be listed as authors because they cannot take responsibility for the accuracy, integrity and originality of published work.

Similarly, COPE guidance emphasises transparency when AI tools are used in academic writing and publication.

These standards are especially important in scientific research because inaccurate AI-generated references or fabricated information can undermine research integrity.

Intellectual property ownership and attribution disputes

AI-assisted outputs may involve multiple contributors, including the user, software developer, organisation, data provider and human editor.

This creates uncertainty concerning ownership and attribution.

For example, a company may use an AI platform to generate advertising material. The platform's terms of service may regulate the contractual relationship between the company and the AI provider.

However, contractual permission to use an output does not automatically establish that the output qualifies for copyright protection.

Important legal considerations include:

1. Originality: Whether the final work contains sufficient original human expression.

2. Human contribution: Whether a person exercised meaningful creative control over the final output.

3. Training data: Whether protected material was used unlawfully in training or generating the output.

4. Licensing: Whether the AI platform's contractual conditions permit the intended commercial use.

5. Attribution: Whether contributors and relevant sources have been appropriately acknowledged.

6. Infringement: Whether the final output reproduces protected expression belonging to another creator.

AI attribution and data protection

AI-generated outputs may also involve personal information, photographs, voice recordings and other identifiable data.

The Digital Personal Data Protection Act, 2023, establishes India's framework for processing digital personal data.

Where personal information is used in generating AI-assisted content, organisations must consider applicable lawful-processing requirements and safeguards.

Attribution alone does not establish permission to use someone's personal data, image or voice.

Similarly, identifying an AI system as the source of content does not remove legal responsibility for privacy violations, unlawful disclosure or misleading representations.

Important case laws and judicial decisions

1. Naruto v. Slater (2018)

United States Court of Appeals for the Ninth Circuit, 888 F.3d 418 (9th Cir. 2018)

The dispute concerned photographs taken by a crested macaque using a camera belonging to a wildlife photographer.

The court considered whether an animal could bring a copyright infringement claim under the US Copyright Act.

The Ninth Circuit concluded that the monkey lacked statutory standing to bring the claim under the relevant copyright legislation.

Legal significance: This decision is relevant to AI attribution because it highlights the importance of identifying a legally recognised author under copyright legislation. Although the court did not decide the copyright status of AI-generated works, the case illustrates the difficulties associated with recognising non-human creators as copyright holders.

2. Thaler v. Perlmutter (2023)

United States District Court for the District of Columbia, 687 F. Supp. 3d 140 (D.D.C. 2023)

The case concerned a copyright application for an artwork titled A Recent Entrance to Paradise, which was claimed to have been generated autonomously by an AI system called DABUS.

The US Copyright Office rejected the application because the work lacked human authorship.

The District Court upheld the Copyright Office's position and concluded that human authorship is a fundamental requirement under US copyright law.

Legal significance: This is one of the most important judicial decisions concerning AI-generated outputs. It establishes that, under the US Copyright Act, a work generated entirely by AI without human creative authorship does not qualify for copyright protection.

3. Feist Publications, Inc. v. Rural Telephone Service Co. (1991)

United States Supreme Court, 499 U.S. 340 (1991)

The dispute concerned the copyrightability of telephone directory information.

The Supreme Court held that copyright protection requires originality involving independent creation and at least a minimal degree of creativity.

The Court rejected the argument that the effort or expense involved in collecting information automatically creates copyright protection.

Legal significance: The judgment is relevant to AI-assisted outputs because the mere generation or compilation of large amounts of information does not automatically establish originality. Human creative contribution and original expression remain important considerations.

4. University of London Press Ltd. v. University Tutorial Press Ltd. (1916)

Chancery Division, [1916] 2 Ch 601

The case concerned copyright protection for examination papers prepared by university examiners.

The court considered the meaning of originality and explained that copyright does not require novelty in the patent-law sense. Instead, a work must originate from its author and involve sufficient independent effort and skill.

Legal significance: This decision provides a foundational understanding of originality in copyright law. It is relevant to AI-assisted writing because human editing, arrangement and independent intellectual contributions may support copyright protection even when AI tools are used during the creation process.

5. Eastern Book Company v. D.B. Modak (2008)

Supreme Court of India, (2008) 1 SCC 1

The case concerned copyright protection claimed over judicial judgments published in edited legal reports.

The Supreme Court examined whether editorial contributions such as headnotes, paragraph numbering and other additions satisfied the originality requirement.

The Court adopted a standard requiring a minimal degree of creativity rather than mere labour or mechanical effort.

Legal significance: This judgment is particularly important for AI-assisted outputs in India. It indicates that the mere application of effort or technical processing is insufficient to establish copyright originality. Human creative skill, judgment and intellectual contribution may be necessary to support protection.

Attribution responsibilities of AI developers and users

Attribution responsibilities may differ according to the role played by each participant.

AI developers: Developers should provide clear information about the capabilities and limitations of their systems and establish appropriate contractual and licensing conditions.

AI users: Users should disclose significant AI assistance when required by academic, professional or contractual standards. They must also verify the accuracy and legality of generated material.

Publishers: Publishers should establish transparent policies governing AI-assisted submissions, authorship declarations and the use of AI-generated images or text.

Research institutions: Institutions should develop guidelines for responsible AI use, plagiarism prevention and research accountability.

Commercial organisations: Businesses should examine copyright, licensing, confidentiality and consumer protection issues before publishing AI-generated material.

Regulatory challenges and emerging legal developments

Several challenges continue to affect the regulation of attribution for AI-assisted outputs.

First, there is no universally accepted international standard defining the minimum level of human creative contribution necessary for copyright protection.

Second, AI systems may generate similar outputs for different users, creating disputes concerning originality and ownership.

Third, identifying the origin of AI-generated material can be difficult because generative systems do not always provide reliable source-level attribution.

Fourth, the use of copyrighted material in AI training raises questions concerning reproduction rights, licensing and exceptions under national copyright laws.

Fifth, academic and commercial organisations require consistent disclosure policies to distinguish genuine human authorship from automated content generation.

The European Union AI Act, 2024, also introduces transparency obligations for certain AI-generated content, including requirements relating to the disclosure and marking of artificially generated or manipulated content in specified circumstances.

These obligations should be distinguished from copyright authorship and ownership rules.

Conclusion

Attribution for AI-assisted outputs is an important emerging issue in intellectual property law, academic integrity and digital governance. The increasing use of generative AI has challenged traditional concepts of authorship, originality, ownership and creative responsibility.

In India, the Copyright Act, 1957, particularly Section 2(d)(vi), provides an important statutory basis for addressing computer-generated works. However, its application to modern generative AI systems requires further judicial and legislative clarification.

Judicial decisions such as Thaler v. Perlmutter, Naruto v. Slater, Feist Publications v. Rural Telephone Service and Eastern Book Company v. D.B. Modak provide important principles concerning legal authorship, originality and human creative contribution.

Ultimately, effective attribution governance requires transparency regarding AI involvement, recognition of genuine human contributions, protection of intellectual property rights and clear accountability for published outputs. A balanced legal framework should encourage responsible AI innovation while protecting creators, researchers, publishers and the public.