Artificial Intelligence and Intellectual
Property Law: Authorship, Ownership, and
Liability in the Age of Generative AI
Author:Abdul Waheed Asghar
LLB (Hons), Punjab University Email: abdulwaheedasghar@[Link]
Abstract
The proliferation of generative artificial intelligence (GenAI) has fundamentally challenged the
anthropocentric foundations of intellectual property (IP) law. This paper examines the
burgeoning doctrinal tension between traditional authorship requirements and autonomous
algorithmic production, filling a critical gap in the literature regarding the lack of a structured,
tiered protection model. Through a comparative doctrinal analysis of the US, UK, and Pakistan,
the study identifies a "protection vacuum" created by the rigid application of human-centric
standards. The research proposes a Hybrid Governance Model—comprised of mandatory
transparency, tiered sui generis protection, and a shared liability framework. This framework
contributes to the field by providing a pragmatic middle-ground that balances the incentive for
technological innovation with the preservation of human creative dignity.
Keywords: Artificial Intelligence, Intellectual Property, Copyright, Generative AI, Thaler v
Perlmutter, OSCOLA, Legal Reform.
1. Introduction
The evolution of artificial intelligence from a passive tool to a generative force has disrupted the
established understanding of the "author" as a biological entity.¹ As GenAI systems produce
sophisticated literary and artistic works with decreasing human intervention, the legal community
faces a fundamental trilemma: the determination of authorship, the allocation of ownership, and
the attribution of liability.²
This paper suggests that maintaining a strict human authorship requirement, without
complementary protection mechanisms, may risk creating significant legal uncertainty for
AI-generated works. To address these shifts, this research adopts specific categorical definitions
to distinguish varying degrees of autonomy.
1.1 Definitions
● AI-Assisted Work: Works where AI is utilized as a functional tool, but the human creator
maintains substantive control over the expressive elements and creative direction.
● AI-Generated Work: Works produced with minimal human creative input, where the
output is primarily determined by the internal parameters of the algorithm.
● Algorithmic Work: Fully autonomous outputs generated without a specific
human-directed prompt or iterative intervention.
2. Literature Review: Thematic Perspectives and Research Gaps
The current academic discourse regarding AI and IP is largely polarized between philosophical
and utilitarian justifications.
2.1 Traditional Authorship and Personality Theory
Traditionalists maintain that copyright is an inherently human right, functioning as a legal
extension of the author’s personality.³ However, this approach is limited in that it struggles to
account for the economic value produced by AI systems that lack biological personhood but
satisfy the functional requirements of "originality."⁴
2.2 Utilitarian and Economic Approaches
Progressives advocate for an "innovation-first" approach, suggesting that AI outputs should be
protected to incentivize technological development.⁵ Yet, a notable critique of this theory—which
often treats AI as a "reasonable robot"—is that it arguably overlooks the moral and economic
rights of human creators whose works are frequently argued to be ingested as training data
without explicit licensing.⁶
2.3 Identification of the Research Gap
While existing literature explores the "all-or-nothing" approach to AI authorship, there is a
distinct lack of a normative framework that scales legal rights according to the level of human
intervention. This paper addresses this void by proposing a tiered protection system.
3. Methodology
This research employs a mixed doctrinal methodology to move beyond purely theoretical debate
and ground its normative proposal in existing legal practice across jurisdictions. The
methodology includes:
● Doctrinal Legal Analysis: Evaluating international treaties (Berne Convention, TRIPS)
and national statutes to determine the scope of current authorship definitions.
● Comparative Jurisdictional Review: Analyzing the diverging paths of the US
(precedent-driven), the EU (regulatory-focused), and Pakistan (emerging framework).
● Case Law Analysis: A critical review of recent appellate decisions, specifically Thaler v
Perlmutter (2025).
● Normative Framework Construction: Formulating a Hybrid Governance Model that
aligns technological reality with established legal principles. This mixed methodology
allows the paper to move beyond theoretical abstraction and ground its normative
proposal in trans-jurisdictional legal practice.
4. Authorship and the Judicial Response
For decades, judicial consensus in jurisdictions that have addressed the issue has been clear:
non-human entities lack the standing to hold intellectual property.⁷ This principle was famously
reinforced in Naruto v Slater, which rejected copyright claims for non-human biological entities.⁸
Recent judicial decisions, including Thaler v Perlmutter (2025), have brought this debate to a
modern crisis point.⁹ In affirming the rejection of Stephen Thaler’s "Creativity Machine" as an
author, the court reinforced the requirement of a "human mind" for copyright eligibility.¹⁰ While
legally sound under current statutes, this ruling could potentially lead to a "protection vacuum"
for works that possess market value but lack biological origins.¹¹ Consistent with the thesis of
this paper, I suggest that this void necessitates a new category of sui generis protection to
prevent a lawless creative commons.¹²
5. Ownership of AI-Generated Works
The UK’s Copyright, Designs and Patents Act 1988 (s 9(3)) credits authorship to the person
who made the "necessary arrangements" for the work.¹³ While this was once a sufficient
standard, it is often argued to be inadequate for modern Large Language Models (LLMs) where
creative control is frequently split between the developer and the user.¹⁴ This paper suggests
that ownership should be a derivative of the creative control exercised—assessed through
prompt specificity, iterative input, and demonstrable human contribution—potentially requiring a
split-title approach in cases of complex co-creation.
6. Liability for AI Infringement
A critical frontier in IP law involves the allegedly unauthorized use of copyrighted works for
model training. AI systems are often argued to contain protected content ingested without
explicit licensing.¹⁵ While developers frequently claim "Fair Use" or "Text and Data Mining"
(TDM) exceptions, this remains an unsettled legal area.¹⁶
A Shared Liability Model suggests a partition of responsibility: developers should bear
"structural liability" for the data curation process, while users should face "output liability" for
specific prompts that intentionally replicate protected works.¹⁷ This approach seeks to avoid the
"all-or-nothing" trap that currently plagues AI litigation.
7. The Pakistani Perspective: A Reformative Opportunity
Pakistan is currently at a legislative crossroads. While the Copyright Ordinance 1962 does not
explicitly address digital autonomy, the Islamabad AI Declaration (2026) signals a move
toward a "Human-in-the-Loop" (HITL) regulatory framework.¹⁸ Analogous principles in land
revenue law suggest that the same standards of "title certainty" required for immovable property
should arguably be applied to "digital title" to ensure market stability.¹⁹ To remain competitive,
Pakistan may need to consider transitioning its IP laws to recognize AI-assisted works, providing
the legal security that international investors demand.²⁰
8. Proposed Legal Framework: The Hybrid Model
To bridge the gap between innovation and the protection of human dignity, this paper proposes
the following legal justifications for a hybrid approach:
● Mandatory AI Attribution: Commercial works should disclose AI involvement to ensure
consumer transparency and maintain the integrity of human creative markets.
● Tiered Sui Generis Protection: Human-only works retain full copyright, while
AI-generated works receive a shortened 20-year term to encourage rapid public domain
entry while protecting initial investment.
● Compulsory Licensing for Training: A statutory framework to compensate human
authors whose works are used for training, thereby balancing developer needs with
creator rights.
9. Conclusion
This research has identified a widening doctrinal failure where the "human spark" requirement
leaves vast amounts of economic value in legal limbo. In response to the central research
question, this paper concludes that the law may need to consider transitioning from treating AI
as a mere tool to recognizing it as a unique category of creative agent. By adopting a hybrid
framework—incorporating transparency and shared liability—jurisdictions can ensure the law
remains a catalyst for progress. While the US continues to rely on a case-law-driven approach
and the EU moves toward a rigid regulatory framework, Pakistan has the opportunity to draft a
flexible, development-oriented code that balances these two extremes. There is an increasing
need for proactive legislative frameworks to ensure technology grows in a settled legal
environment.
Footnotes (OSCOLA Style)
1. Ryan Abbott, The Reasonable Robot: Artificial Intelligence and the Law (CUP 2020) 12.
2. WIPO, ‘Revised Issues Paper on Intellectual Property Policy and Artificial Intelligence’
(WIPO/IP/AI/2/GE/20/1, 2020).
3. Jane C Ginsburg, ‘The Concept of Authorship in Comparative Copyright Law’ (2003) 52
DePaul L Rev 1063, 1068.
4. Ginsburg (n 3) 1070.
5. Abbott (n 1) 45.
6. Abbott (n 1) 56.
7. Naruto v Slater 888 F.3d 418 (9th Cir 2018).
8. ibid.
9. Thaler v Perlmutter No 23-5233 (DC Cir Mar 18, 2025).
10.US Copyright Office, ‘Copyright Registration Guidance: Works Containing Material
Generated by AI’ (2023) 88 FR 16190.
11.Ginsburg (n 3) 1082.
12.See Section 8 above.
13.Copyright, Designs and Patents Act 1988, s 9(3).
14.Andres Guadamuz, ‘Artificial Intelligence and Copyright’ (2017) WIPO Magazine.
15.WIPO Report (n 2) 5.
16.Authors Guild v Google, Inc 804 F.3d 202 (2d Cir 2015).
17.Pamela Samuelson, ‘Allocating Ownership Rights in Computer-Generated Works’ (1985)
47 U Pitt L Rev 1185, 1201.
18.Ministry of Information Technology, ‘Islamabad Declaration on AI Ethics’ (9 February
2026).
19.Punjab Land Revenue Act 1967 (contextual comparison regarding titles).
20.IPO-Pakistan, ‘National IP Strategy Review’ (March 2026).