Introduction
The creation and development of the legal regime applicable to copyright, related rights and sui generis rights was influenced by constant technological developments. The new and dynamic generative AI technology raises new opportunities and challenges for the current law and rightsholders.
There are rich ongoing research and debates that address the issue of the magnitude of the potential impact on the current legal frame, brought by an unprecedented scale of automation that increase productivity and access to creativity, that raises legal uncertainty and questions of how can and should EU law meet challenges brought by generative AI.
Throughout this Study, the terms”generative AI” and ”AI system” are understood as ”a machine-based system that is designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments”, according to the definition in Art. 3 of the Regulation (EU) 2024/1689 of June 13, 2024, laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)[1].
The word ”copyright” will be used as a translation of the term ”author's rights”, this being the wording adopted by the European legislator in EU law. From a legal point of view, the two terms are not equivalents, on the contrary, there are many differences between the legal regime applicable to copyright in common law systems and to author's rights in continental, civil law systems. For example, sound recordings and broadcasts are protected under copyright in the United Kingdom, a common law country[2], while the same are protected under a related or neighbouring right in continental, civil law countries and EU legislation[3].
The Study is organised in three parts, followed by a short conclusion. The first part raises the question if works created with the use of generative AI might be protected under copyright, analysing the principles of human authorship and the basic requirement of originality, leading to the conclusion that only human authors may acquire copyright protection under current legislation. The second part addresses the issue if the existing related rights may protect AI-generated works, in view of the criteria for acquiring legal protection. The third part puts forward and analyses the policy option for protection of non-copyrightable AI-generated outputs by a newly enacted and carefully conceived related right, granted to legal persons that develop generative AI systems.
1.1. The principle of human authorship under copyright
Intellectual property law (comprising industrial property law, copyright and related rights) is conceived and build on the fundamental principle of human authorship. Only a human being, a natural person may create works that do not exist in nature as such and that are the expression of a ”self”, thus acquiring the quality of ”author” under copyright law.
In copyright law, the principle of human authorship is rooted in key legal provisions of public international law, European Union law and national legislations[4].
The Berne Convention for the Protection of Literary and Artistic Works of September 9, 1886[5] (”the Berne Convention”) is the cornerstone and the common ground for the protection of copyright, that sets the minimum level of legal protection in contracting states.
The main reasons for adopting the Berne Convention is to encourage and economically reward the creativity of human beings, the dissemination of works of human creativity and their legitimate use by third parties.
Article 1 of the Berne Convention states that protection is granted to ”rights of authors” in their ”literary and artistic works” - a term widely interpreted in art. 2 (1) of the Convention. Author's rights are granted by the mere fact of creation of the work. ”Authorship” of a work is a legal presumption, based on the mere publication a work under a personal name, with the ”consent” of the author [article 3 (3) of the Berne Convention]. ”Consent” for publication of a work may be given only by a human author.
Copyright comprises personality rights or moral rights[6] and economic rights [art. 6bis (1) of the Berne Convention]. Only human authors may be granted personality rights or moral rights, such as the right to claim authorship of a work.
The term of protection of a work under copyright is calculated based on the ”life of the author” and ”his death” [art. 7 (1) of the Berne Convention], which leads to the conclusion that copyright was created for human beings only.
The WIPO Copyright Treaty of December 20, 1996[7] (”the WCT”) reaffirms the existing obligations undertaken by Contracting Parties under the Berne Convention and, recognising ”the profound impact of the development and convergence of information and communication technologies on the creation and use of literary and artistic works”[8] provides for the protection of computer programs, as literary works, under the Berne Convention and of compilations of data (databases), as such.
The European Union is a party to the WIPO Copyright Treaty and is thus obliged to comply with the Berne Convention. European Union copyright law and national rights of EU Member States are constructed on the basis of the minimum protection granted under the Berne Convention and the WIPO Copyright Treaty to works of human authors.
The legal regime of author's economic rights in the EU is harmonised by the Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the harmonisation of certain aspects of copyright and related rights in the information society[9] (”the Infosoc Directive”), which implements the WIPO Copyright Treaty in the EU[10].
Moral rights of the author remain under the legal regime established by international law (the Berne Convention and the WIPO Copyright Treaty) and by national law of each EU Member State.
One of the conditions for classifying a subject matter as a ”work”, protectable by copyright under the Infosoc Directive, is that the work is ”the author’s own intellectual creation”[11]. There is no work without an author and such author has to be always a human being, a natural person. There always has to be a direct link between the subject matter, or the creation, or the work and the intelligence of a natural person, a human being, in order to protect it under copyright[12].
The copyright term of protection of is harmonised at the EU level by the Directive 2006/116/EC of the European Parliament and of the Council of December 12, 2006 on the term of protection of copyright and certain related rights[13] (the ”Term Directive”) and it is established for ”the life of the author and for 70 years after his death” [art. 1 (1) of the Term Directive]. Copyright may be granted only to an initial human author, for the duration of the life of a human being, plus 70 years after the death of a human being.
In United Kigdom (a common law country, former member of the European Union), copyright protection is granted to computer-generated works, defined by law as works ”created in circumstances such that there is no human author of the work” [sec. 178 of the UK Copyright, Designs and Patents Act 1988[14]].
The author of a computer-generated work is taken to be ”the person by whom the arrangements necessary for the creation of the work are undertaken” [sec. 9 (3) of the UK Copyright, Designs and Patents Act 1988]. There is no further qualification of the term ”person” as natural or legal person.
The term of copyright protection of computer-generated works is 50 years ”from the end of the calendar year in which the work was made”, without reference to the life or the death of the author as a natural person [sec. 12 (7) of the UK Copyright, Designs and Patents Act 1988].
1.2. ”Originality” of a work, as a fundamental requirement for copyright protection under European Union law
”Originality” is one of the fundamental, sine qua non, requirement for the legal protection of a work under the continental author's right (not in all cases under copyright, in common law).
The legal term ”originality” is not defined by the Berne Convention, who makes reference to the ”original work”, in the sense of the initial work, originating from his author [art. 1 (3) and 14bis (1) of the Berne Convention].
This understanding of the legal term ”original” lead in common law legal systems to the interpretation of ”originality” in an objective way: a work is original if it is not a copy of another work. At the heart of copyright is the exclusive right of the rightowner to permit or prohibit third parties to copy his/her original (initial) work.
The United Kingdom copyright legislation was, up to a certain level, harmonised with the ”originality” standard established by the CJUE jurisprudence.The public consultation of United Kingdom Intellectual Property Office (”UKIPO”) on Artificial Intelligence and IP states that ”for a work to be original it must be the author's own intellectual creation. This means that the author has made free and creative choices and the work shows his personal touch”[15].
Under EU law and continental, civil law legal systems, ”originality” is assessed in relation to an object matter that must be an ”expression” of author's ”self”[16] or ”personality”. Only human beings are endowed with a ”self” and a ”personality” that may find an expression in works created by human intelligence.
According to article 2 of the WCT: ”Copyright protection extends to expressions and not to ideas, procedures, methods of operation or mathematical concepts as such”. Algorithms are thus excluded from copyright protection. So are mere instructions or prompts given by a natural person to an AI system.
In consistency with the WCT, the jurisprudence of the CJUE specifies that the concept of ”work” is understood as the ”expression” of ”the author’s own intellectual creation”[17], that is, an intellectual creation of a natural person[18].
In the most recent Mio judgment[19], the CJUE restated that ”for the concept of ‘originality’ and, consequently, for copyright protection under EU law”, ”the requirement that the author’s personality be reflected in the subject matter for which protection is claimed, through the expression of that author’s free and creative choices, is decisive”.
The making of a ”free and creative choice” requires the rational capacity to counsciusly decide with regard to the creative elements of the work, in a free manner. Such capacity of discernment is caracteristic of human intelligence only.
A work which is the result of ”free and creative choices”, or of a ”creative freedom” means that its realisation has not been dictated ”by technical considerations, rules or other constraints, which have left no room for creative freedom”[20].
Where the expression of a subject matter is dictated by its technical function, ”the different methods of implementing an idea are so limited that the idea and its expression become indissociable”[21], that leading to the exclusion of copyright protection.
1.3. The possibility of granting copyright in AI-generated works to natural persons
In view of the current EU copyright legal regime, copyright may not be denied to a work, merely because an AI system was used for its creation.
Copyright may not be granted to developers of AI systems, because legal persons may not be qualified as ”authors” of a work. Developers of an AI system may own copyright in the AI software, but that does not provide them copyright in the output of the AI software[22].
Only natural persons may have the quality of ”authors”, able to initially acquire copyrights in ”AI-generated works”[23], by the mere fact of the creation of a work, by using AI as a tool. Currently, such AI-generated works, potentially protectable under copyright are music[24], paintings[25], images, press articles[26], movies[27], novels etc.
The main question in deciding to recognise copyright in such AI-generated outputs is the same with regard to any type of human created works: does the work fulfil the criteria of ”originality”? Is the work ”original”, in the sense that it is the expression of free and creative choices that reflect the personality of a human author? Or are such expressions dictated by technical constraints imposed by the algorithms of the AI system?
In order to ”reflect the personality of a human author”, the choices, such as instructions or prompts given to the AI system, have to be ”creative” enough to be expressions of personality, not mere ideas, which are excluded from protection. As decided in the Mio case, ”the use of the words ‘reflects’ and ‘expression’ clearly indicates that such choices and the author’s personality must be visible in the subject matter for which protection is claimed (our emphasis)”[28], namely, in the AI-generated work or output itself. The work protected by copyright must be ”expressed in an objective manner”, ”clearly and precisely identified”, in order to ensure legal certainty[29].
In order to be ”free”, the creative choices must be conscious decisions of the human author, that are not determined by technical constraints of the AI system. Such technical constraints are the algorithms of the AI system and also the set of data used for the training of AI[30].
AI algorithms and training data sets do determine the limits of human choices, such as instructions or prompts given to the AI system. The algorithms of an AI system may not be controlled by human prompts. Random algorithmic choices may break the link between the human choice and the generated work[31]. Training data sets leave no room for human imagination able to surpass and transcend what already exists and is already ”known” by the AI system.
One of the main risks in granting copyright protection to AI-assisted works is false attribution or false representation that a work generated by AI with minimum human intervention that lacks any originality is the result of human free creative choices, in order to benefit from the term of copyright or author's rights protection (life of the author plus 70 years after his death). Another negative result is the subsequent restriction of use of such AI-assisted works by third parties, including other AI developers, which is certainly able to stifle further creativity because of scracity of new training data.
This risk is mitigated under the CJUE jurisprudence, according to which where the author has made choices that are not dictated by technical or other constraints, it cannot be presumed that those choices are creative, for the purpose of acquiring copyright protection. As decided in the Mio case, creative choices in a work must be identified: ”the court before which the question of the originality of a utilitarian object is brought must seek out and identify the creative choices in the shape of that subject matter in order to be able to declare it protected by copyright”[32].
In accordance with the CJUE jurisprudence, the Munich District Court stated, in its decision[33] of February 13, 2026: ”1. Whether products generated by artificial intelligence have the character of a work within the meaning of Section 2 Paragraph 2 of the German Copyright Act depends on the extent to which human creative influence is still exerted, despite the software-controlled process. 2. Copyright protection is therefore conceivable as a result of human intervention in AI results, which can also take place subsequently or successively during prompting and which leads to the personality of the prompter being reflected in the output. 3. However, human influence must shape the resulting output in a sufficiently objective and clearly identifiable way. This is the case, at least, when the creative elements incorporated into the prompting process so dominate the output that the object as a whole can be considered the author's own original creation”.
The person claiming copyright protection in AI-generated works bears the burden of proof of the existence of a creative intellectual creation, namely, of the fact that an ”independent creative expression of his personality is discernible” in the AI-generated work.
The effort to formulate and test a complex prompt of at least 1700 characters to create a work, characterized solely by time expenditure, is not a criterion for assessing originality.
It is essential that the AI model is ”closer to an aid than to an independent creative instrument". The AI's technical activity must not outweigh the user's creative and artistic influence on the output (para. 26 and 27).
In September 2025, Italy adopted the Law no. 132/2025 on the protection of AI[34], the first national law on AI. According to the Law, ”intellectual works” must be of ”human origin”. Art. 25 of the Law expressly reserves copyright for human-created works, including those made by using tools of AI, if they are the result of human authors' intellectual efforts[35]. The Law reaffirms copyright protection of works according to provisions of the Berne Convention.
The same approach was adopted by the Municipal Curt of Prague, in its Decision[36] of June 20, 2023, in the case S. Š. v TAUBEL LEGAL, advokátní kancelář s.r.o.. The Prague Court ruled that the person who claims authorship of AI-generated works and copyright in the work, based on his prompts or instructions given to the AI system, has the obligation to adduce specific evidence in support of such claims (para. 11). An AI-generated work is not a work created personally by the author (the user of the system) himself, but with the help of AI, according to specific prompts that do not constitute works of authorship, but mere ideas, excluded from copyright protection (para. 12).
The US Copyright Office adopted the same approach to the copyrightability of AI-generated outputs, in its ”Report on Copyright and Artificial Intelligence”, Part 2, of January 29, 2025[37].
The Office reaffirmed that copyright protection requires human authorship and that, according to an established jurisprudence: “copyright has never stretched so far [as] ... to protect works generated by new forms of technology operating absent any guiding human hand”[38].
”Prompts essentially function as instructions that convey unprotectible ideas. While highly detailed prompts could contain the user’s desired expressive elements, at present they do not control how the AI system processes them in generating the output”[39].
Copyright protection may be available to AI-generated works only if AI was used as a tool for human creativity, insofar as there is ”sufficient human control” over the expressive elements of the work.
The current functioning of AI systems shows gaps between prompts and the resulting outputs[40] and demonstrate that ”the user lacks control over the conversion of their ideas into fixed expression”, and that ”AI systems are largely responsible for determining the expressive elements in the output. In other words, prompts may reflect a user’s mental conception or idea, but they do not control the way that idea is expressed”[41]. Only expressions may be copyright protected, ideas being expressly excluded from protection.
Copyright may be granted to human authors, users of the AI system, if their works of authorship are ”perceptible in the AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs”[42].
Korea, Japan, China also apply the principles that only natural persons may be authors of a work and that copyright protection requires proof of a human creative contribution to the expressive form of the work[43].
Rights related to copyright were enacted due to technological developments, for the benefit of performers, producers of phonograms and broadcasting organisations, under the International Convention for the Protection of Performers, Producers of Phonograms and Broadcasting Organizations done in Rome, October 26, 1961[45] (the ”Rome Convention”), followed by the WIPO Performances and Phonograms Treaty of December 20, 1996 (the ”WPPT”).
Recently, the Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market (the ”DSM Directive”) enacted related rights to press editors (legal persons), for the online use of their press publications by information society service providers[46].
The grant of legal protection under related rights is not based on the principle of ”human authorship”. Consequently, related rights are not granted in view of the ”originality” of phonograms, broadcasts or press publications. The criterion for protection is a mere economic investment in the production of the subject-matter (except for performances).
That allows legal persons, such as phonograms and videograms producers, broadcasters and press publishers to acquire related rights in respect of their phonograms, videograms, broadcasts and press publications, including those created with the use of AI systems and even if there is no human creative contribution in the process[47].
An extensive study on AI music outputs[48] concludes that phonogram producers benefit from protection under a related right regardless of whether the underlying sound was created by a human or AI-generated. There is no threshold for protection other than the qualification of the subject matter as a phonogram[49]. The fixation requirement implies that AI-enabled continuous unfixed music generation and streaming projects are not phonograms and hence are not covered by rights awarded to phonogram producers.
The study gives examples of broadcasting events involving songs created with AI that already took place, such as the AI Song Contest, organised by the Dutch national public broadcaster, with a first edition in 2020 and of web radios broadcasting AI-generated music.[50]
The conclusion of the study is that the use of AI systems in music production does not impact on the legal qualification of protection of broadcasts by related rights. ”This type of related rights could be applicable to musical content that, due to the lack of fixation, is not protected by copyright and/or related rights of phonogram producers.”
The authors stress that when assessing the desirability of proposals for new modes of protection of AI outputs, the availability of related rights protection for non-copyright protected AI-generated outputs should be considered[51].
Under the current copyright legal regime, AI-generated works, such as music, films, novels, pictures, software etc., may receive copyright protection granted only to human authors, if they are created with a prevailing human original or creative contribution, that can be discernable or perceptible in the AI-generated output.
If the copyright criteria for protection are not met, AI-generated works fall into the public domain, with the risk to undermine investment and to raise competition concerns[52].
AI-generated works such as phonograms, broadcasts and press publications may be protected by related rights, granted to legal persons (producers, press publishers), based on the criteria of economic investments in the production process.
In view of the above legal analysis, we support the opinion that the current copyright (author's rights) and related rights legal regime may not be stretched to cover AI-generated outputs that does not meet the established criteria for protection[53].
We put forward the question if there is a need to protect investments in the development of AI systems and to incentivize further investment by enacting a new related right for the protection of non-copyrightable AI-generated outputs, based on the criteria of economic investments in the several stages of the creation of an AI system[54]? As pointed by reputed authors, ”training algorithms vary in originality from standard to uniquely developed. Many established training algorithms are standard, and available online in open-source libraries in the form of pre-written software. However, for some problems, new algorithms have to be developed, necessitating human and financial investments [...]”[55]. Prompt engineering, iterative curation, and fine-tuning of models may involve substantial expertise, as argued in the study on ”Generative AI and Copyright. Training, Creation, Regulation”[56].
Should these acts be treated as acts of creation of the AI-generated output (involving a much higher level of creativity that users' prompts) or as a technical manipulation? While some authors argue that the natural persons involved in the construction of the AI system, such as engineers, coders, data trainers, may be considered the co-authors and co-owners of copyright of the AI-generated outputs in cases where there is no creative contribution from the user of the AI system[57], other authors sustain, on the contrary, that in such cases the persons that created the AI system may not be considered authors of the output, as they have no influence or control on the final result, which is random[58].
Should fully AI-generated outputs, with no human creative intervention[59] able to attract copyright protection, remain in the public domain or should AI developers (legal persons) be granted a limited protection under a new related-right?
These are policy question that take into consideration not only the societal and economical value of AI systems, the investments in creating such value, that would justify the grant of exclusive rights in order to allow their recoupment and to prevent market failures[60], but also the economic and social risks for granting protection of AI-generated works under a related right.
The recent study on ”The economic of copyright and AI - Empirical evidence and optimal policy”[61], requested by the JURI Committee of the European Parliament shows that ”the amortised cost of training top foundation models exceeds tens of millions of dollars. These costs have been increasing at a rate of approximately 240% per year since 2016, and have grown so large that, with a few exceptions of recent government-backed investments, only commercial players can afford to produce frontier models. If this trend continues, the most expensive training runs could exceed $1B by 2027.”[62]
A newly enacted related right for the protection of AI-generated outputs[63] would allow the recoupment of economic investments and would prevent losses through acts of free use and free riding by third parties[64].The new related right would be granted to AI developers for AI-generated works that do not meet the threshold of copyright protection[65] and may not benefit from protection under the existing related rights, based on the criterion of economic investment[66].
Such normative intervention addresses legal uncertainties created by difficulties in establishing and proving the limits between AI and human creative contribution to output[67] and, in view of the legal presumption of authorship, discourages circumvention of copyright by false claims of human authorship[68].
For AI developers it is difficult to prove how the AI functions: as a mere tool controlled by human beings who make creative choices in the process of generating an output or as an automated process, capable to determine an expressive outcome with no human creative contribution[69]. ”Even though the output is in principle deterministic and traceable, it is often not human-explainable due to the complexity of the calculations, especially in the case of artificial neural networks (the ‘black box’ issue)”[70].
For AI-users it is difficult to prove that their contribution is, first, ”sufficient” to outweight the algorithms of the AI system and to control the output and, second, ”creative”, so that it confers ”originality” on the output as a whole. As explained in the US Copyright Office Report, Part 2[71]: ”Although AI technology continues to advance, uncertainty around how a particular prompt or other input will influence the output may be inherent in complex AI systems built on models with billions of parameters”. Issues related to technical solutions for the identification of AI-generated works were addressed by recent legislative proposals and several solutions were envisaged[72].
Another crucial argument for granting new related rights in AI-generated outputs is the liability for AI-generated works, given that the technological developments lead to more autonomous AI systems, capable ”not only of creating, but of initiating tasks, selecting data inputs, and refining outputs”[73].
A careful analysis of economic data is required in the eventual conception of a new related right[74]. The scope of protection under the new related right should be limited and its the term of protection should be shorter, to allow the recoupment of investments in the develpment of AI systems, while preventing obstacles in further obtaining AI-generated works based on newly created and publicly available data. When establishing the duration of the new right, one might have in view the 2 years term of protection of the press publishers related rights [Art. 15 (4) of the DSM Directive[75]].
However, as argued by reputed authors, one should observe the ”market dynamics in relation to the AI output per se” and bear in mind that if the market dynamic is such as to minimize the lifecycle of AI outputs, ”once the substitution of AI products or services happens with such speed that investments could not be recouped despite IP protection in place, any IP right would be detrimental to economic welfare and not justifiable.”[76]
Against the enactment of a new related right for AI-generated outputs is the expansion of monopoly rights over AI-generated outputs, to the detriment of the public domain and the scarcity of new data allowing further AI development[77].
Another risk is the encouragement of massive, rapid, low cost of AI production of outputs, that would substitute the demand for original works of authorship, to the detriment of authors. Recent studies show that generative AI can depress creative supply[78]. This would undermine the copyright system and would run against the very scope of author's rights to protect and encourage human creativity and the promotion of culture.
Other authors are of opinion that AI-generated works do not substitute the demand for human works and thus do not disrupt the market of such works, because the AI output only reproduces unprotected ideas, concepts and styles of the human authors[79].
Strong opinions are in the sense that the introduction of a new protection regime (a new related right) for AI-generated output is not justified according to the current state of knowledge and that algorithms and models of an AI system may be protected under extra IP regimes that address market failure, such as trade secrets, unfair competition, technological measures, contract law.[80]
The policy option to enact a new related right was under public consultation in the United Kingdom[81] (policy option 2). The UK legislator asked if the current copyright protection for computer-generated works, generated without any human contribution, should be replaced with a new right of reduced scope and duration. The UK Government response to consultations[82] is that, for the time being, legislative changes lack supporting economic evidence.
The same view was taken by the US Copyright Office[83], given that currently there are not enough empirical data in support of legislative changes.
The European Commission[84] 2020 report emphasizes that “fully autonomous AI-processed outputs currently fall outside the scope of copyright protection”. However, the report further notes that related rights or sui generis mechanisms might eventually be explored for such works, but for now the legal framework remains centered on the human creator.
The recently adopted European Parliament resolution of March 10, 2026 on copyright and generative artificial intelligence - opportunities and challenges[85] underlines that ”K. [...] the development, deployment and use of AI must be fully compliant with the current legal framework [...]”, that ”AE. [...] EU copyright law remains grounded in the principles of human authorship [...]” and that ”AG. [...] international convergence and the establishment of a global regulatory framework [...]” is needed.
In its resolution, the European Parliament insists that content fully generated by AI that does not meet the established criteria for copyright protection should remain ineligible for copyright protection, and that the public domain status of such outputs be clearly determined”.
Conclusion
For the moment, studies, debates, public consulations on a legislative intervention for the enactment of a new type of IP protection (a related right or a sui generis rights) to AI-generated works that may not be protected by copyright arrived at the conclusion that there are currently no sufficient empirical data for such an action[86].
However, they also stressed that the dynamic technological developments in the field should be further monitored[87] and that the enactment of a new type of IP protection of AI-generated output (reled right or sui generis protection) might be an option, if the economical and social evolutions justify such policy option.
[1] Regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828, OJ L, 2024/1689, 12.7.2024.
[2] Sec. 1 ((b) of the UK Copyright, Designs and Patent Act 1988, available at https://www.legislation.gov.uk/ukpga/1988/48/contents, last accesssed on March 17, 2026
[3] For example, under the Directive 2006/116/EC of the European Parliament and of the Council on the term of protection of copyright and certain related rights, available at https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:02006L0116-20111031, last accesssed on March 17, 2026). A comparison between the two legal systems exceeds the scope of this study.
[4] For an analysis of the current national copyright law in the context of AI, in Germany and Belgium, see France Vehar and Thomas Gils, ”I'm sorry AI, I'm afraid you can't be the author (for now)”, in the Journal of Intellectual Property Law and Practice, 2020, Vol. 15, No. 9, pp. 718-726.
[5] Available at https://www.wipo.int/wipolex/en/text/283698, last accessed last accesssed on March 17, 2026.
[6] For a comparative analysis of copyright laws, see Päivi Hutukka, Copyright Law in the European Union, the United States and China, in IIC (2023) 54:1044–1080, https://doi.org/10.1007/s40319-023-01357-0. According to the author, the requirement of the Berne Convention to recognize moral rights was assumed for a long time to be the main reason why the USA did not sign and ratify the Convention until 1989. The USA were reluctant to embrace the concept of moral right of the author.
[7] Available at https://www.wipo.int/wipolex/en/text/295166, last accessed March 16, 2026.
[8] Preamble of the WCT.
[9] Published in the Official Journal L 167 of 22/06/2001 P. 0010 - 0019.
[10] ECJ decision of 13 November 2018, Levola Hengelo BV/Smilde Foods BV, C-310/17, EU:C:2018:899, para. 38.
[11] ECJ decision of 13 November 2018, Levola Hengelo BV/Smilde Foods BV, C-310/17, EU:C:2018:899, para. 36, citing the ECJ judgment of 4 October 2011, Football Association Premier League and Others, C‑403/08 and C‑429/08, EU:C:2011:631, para. 97.
[12] Peter Georg Picht, Florent Thouvenin, ”AI and IP: Theory to Policy and Back Again – Policy and Research Recommendations at the Intersection of Artificial Intelligence and Intellectual Property”, in IIC 54, 916–940 (2023). https://doi.org/10.1007/s40319-023-01344-5.
[13] Published in the Official Journal L 372 of 27.12.2006, p. 12.
[14] Available at https://www.legislation.gov.uk/ukpga/1988/48/part/I/chapter/I, last accessed on March 16, 2026.
[15] Available at https://www.gov.uk/government/consultations/artificial-intelligence-and-ip-copyright-and-patents/outcome/artificial-intelligence-and-intellectual-property-copyright-and-patents-government-response-to-consultation, last accessed March 17, 2026. The UKIPO help six roundtables of public consultations in November and December 2021.
[16] Iaia, Vincenzo, ”To Be, or Not to Be ... Original Under Copyright Law, That Is (One of) the Main Questions Concerning AI-Produced Works” (August 11, 2022). GRUR International, 71(9), 2022, 793-892, Available at SSRN: https://ssrn.com/abstract=4528143. The author shows that ”the exclusive protection of te human creative endeavour finds support form a philosophical perspective too, considering that the etymology of the term 'author' derives from the ancient Greek αὐτός, which means 'self'”.
[17] ECJ decision of 13 November 2018, Levola Hengelo BV/Smilde Foods BV, C-310/17, EU:C:2018:899, para. 36, citing the ECJ judgment of 4 October 2011, Football Association Premier League and Others, C‑403/08 and C‑429/08, EU:C:2011:631, para. 97.
[18] ECJ judgment of 12 September 2019, Cofemel, C‑683/17, EU:C:2019:721, para. 29; ECJ decision of 13 November 2018, Levola Hengelo BV/Smilde Foods BV, C-310/17, EU:C:2018:899, para. 36, citing the ECJ judgment of 4 October 2011, Football Association Premier League and Others, C‑403/08 and C‑429/08, EU:C:2011:631, para. 97.
[19] ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 50.
[20] ECJ judgment of 12 September 2019, Cofemel, C‑683/17, EU:C:2019:721, para. 30-31 and the cited case-law; ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 49 and 63.
[21] ECJ judgment of 11 June 2020, Brompton Bicycle, C‑833/18, EU:C:2020:461, para. 27 and the case-law cited; ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 64.
[22] France Verhar and Thomas Gils, op. cit, p. 721.
[23] Iaia, Vincenzo, op. cit. supra.
[24] Spotify published AI composed music, see ViIaia, Vincenzo, op. cit., note 23.
[25] A Rembrand painting was produced by AI, see ViIaia, Vincenzo, op. cit., note 24.
[26] ”The Guardian” published a press article written by the AI system GPT-3, see ViIaia, Vincenzo, op. cit., note 22.
[27] Netflix released an AI produced horror movie, see ViIaia, Vincenzo, op. cit., note 24.
[28] ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 71.
[29] ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 72.
[30] AI developers are constantly suited for copyright infringement by use of protected works for the training of AI. One recent case is brought by Encyclopaedia Britannica against Open AI: https://www.g4media.ro/enciclopedia-britannica-da-in-judecata-openai-pentru-continutul-utilizat-in-antrenarea-chatgpt.html, last accessed on March 17, 2026.
[31] Iaia, Vincenzo, op. cit. supra.
[32] ECJ decision of 4 December 2025, Mio and konektra, C-580/23 and C-795/23, EU:C:2025:941, para. 65.
[33] Decision in the case 142 C 9786/25, available at https://www.gesetze-bayern.de/Content/Document/Y-300-Z-BECKRS-B-2026-N-1513?hl=true, last accessed March 18, 2026.
[34] Accessible at: https://www.normattiva.it/atto/caricaDettaglioAtto?atto.dataPubblicazioneGazzetta=2025-09-25&atto.codiceRedazionale=25G00143&atto.articolo.numero=0&atto.articolo.sottoArticolo=1&atto.articolo.sottoArticolo1=0&qId=345ab9c5-79ac-4964 a1b1197dd0bdc95f&tabID=0.5253345485394565&title=lbl.dettaglioAtto, last accessed March 20, 2026.
[35] ”Alla legge 22 aprile 1941, n. 633, sono apportate le seguenti modificazioni: a) all'articolo 1, primo comma, dopo le parole: «opere dell'ingegno» è inserita la seguente: «umano» e dopo le parole: «forma di espressione» sono aggiunte le seguenti: «, anche laddove create con l'ausilio di strumenti di intelligenza artificiale, purché costituenti risultato del lavoro intellettuale dell'autore»”.
[36] Decision in the case no. 10 C/2023-16, available at https://msp.gov.cz/documents/14569/1865919/10C_13_2023_10/108cad3e-d9e8-454f-bfac-d58e1253c83a, last accessed March 20, 2026.
[37] Available at https://www.copyright.gov/ai/, last accessed March 20, 2026.
[38] The decision of the the U.S. District Court for the District of Columbia, Thaler, cited in The US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 8.
[39] The US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 18.
[40] With regard to the condition of human intervention in the process of generating works with the use of AI, see also Benhamou, Yaniv; Andrijevic, Ana, ”Intelligence artificielle générative d’images et droit d’auteur”, in Richa, Alexandre; Canapa, Damiano (Ed.) ”Aspects juridiques de l’intelligence artificielle: projet de règlement européen, titularité des données, images et droits d’auteur, droit de la concurrence, responsabilité du fait des produits, prestations de services, Stämpfli Editions SA Berne, 2024, p. 41–85 (Collection lausannoise, CEDIDAC, vol. 98), available at https://serval.unil.ch > resource > REF.
[41] The US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 19.
[42] See as example Rose Enigma, VAu001528922 (March 21, 2023), registered by the Office for copyright protection, in the US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 23, page. See also: https://www.wipo.int/en/web/wipo-magazine/articles/us-copyright-office-on-ai-human-creativity-still-matters-legally-73696.
[43] See, also for examples of copyright policy decisions in other countries, the US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 28-29.
[44] In common law systems, such as United Kingdom, sound recordings and broadcasts are protected under copyright, not under a related right [sec. 1 (1) (b) of the UK Copyright, Designs and Patents Act 1988]. ”Related rights” are designated by the term ”copyright” in common law countries, see explanations at note 1.
[45] Available at https://www.wipo.int/wipolex/en/text/289757, last accessed on March 17, 2026.
[46] Art. 15 of the Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market and amending Directives 96/9/EC and 2001/29/EC, OJ L 130, 17.5.2019, pp. 92–125 (the ”DSM Directive”).
[47] France Verhar and Thomas Gils, op. cit, p. 723.
[48] Oleksandr Bulayenko, João Pedro Quintais, Daniel Gervais and Joost Poort, “AI Music Outputs: Challenges to the Copyright Legal Framework”, reCreating Europe Report (February 2022), available at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4072806, last accessed March 21, 2026.
[49] idem, pages 96-98.
[50] idem, page 100.
[51] idem, page 113.
[52] Nicola Lucchi, Study on ”Generative AI and Copyright. Training, Creation, Regulation”, of July 9, 2025, PE 774.095, page 45, 95, available at https://www.europarl.europa.eu/thinktank/en/document/IUST_STU(2025)774095, last accessed on March 20, 2026.
[53] Péter Mezei, From Leonardo to the Next Rembrandt – The Need for AI-Pessimism in the Age of Algorithms (July 24, 2020). UFITA, 2020, 84(2), p. 390-429. (https://doi.org/10.5771/2568-9185-2020-2-390), available at SSRN: https://ssrn.com/abstract=3592187, last accessed on March 21, 2026.
[54] ”Technical Aspects of Artificial Intelligence: An Understanding from an Intellectual Property Law Perspective”, Drexl, Josef and Hilty, Reto and Beneke, Francisco and Desaunettes-Barbero, Luc and Finck, Michèle and Globocnik, Jure and Gonzalez Otero, Begoña and Hoffmann, Jörg and Hollander, Leonard and Kim, Daria and Richter, Heiko and Scheuerer, Stefan and Slowinski, Peter R. and Thonemann, Jannick (October 8, 2019). Max Planck Institute for Innovation & Competition Research Paper No. 19-13, Available at SSRN: https://ssrn.com/abstract=3465577. last accessed March 20, 2026.
[55] idem.
[56] Nicola Lucchi, op. cit, p. 97.
[57] Jane C. Ginsburg and Luke Ali Budiardjo, ”Authors and Machines”, Berkeley Tachnology Law Journal, Vol. 34:343, pp. 345-445, available at https://btlj.org > articles2019 > 01_Ginsburg_Web, last accessed March 20, 2026.
[58] Benhamou, Yaniv; Andrijevic, Ana, ”Intelligence artificielle générative d'images et droit d'auteur : utilisation des données d'entraînement et protégeabilité des résultats”, In: Arts-Loi// Kunst-Wet: plateforme du droit et de l’art contemporain. Sandrine Carneroli (Ed.). WIELS, Bruxelles. Bruxelles:
Larcier-Intersentia, 2024. p. 81–118. (Création Information Communication), available at https://archive-ouverte.unige.ch/home, last accessed March 20, 2026.
[59] ”As far as “AI-generated output” is concerned, it has to be stressed from the outset that according to the current state of knowledge, really “independently acting” computers do not exist”, see Hilty, Reto and Hoffmann, Jörg and Scheuerer, Stefan, ”Intellectual Property Justification for Artificial Intelligence” (February 11, 2020). Draft chapter. Forthcoming in: J.-A. Lee, K.-C. Liu, R. M. Hilty (eds.), Artificial Intelligence & Intellectual Property, Oxford, Oxford University Press, 2020, Forthcoming, Max Planck Institute for Innovation & Competition Research Paper No. 20-02, Available at SSRN: https://ssrn.com/abstract=3539406, last accessed on March 20, 2026.
[60] Hilty, Reto and Hoffmann, Jörg and Scheuerer, Stefan, ”Intellectual Property Justification for Artificial Intelligence” [...], op. cit.
[61] Christian Peukert, Study on ”The economic of copyright and AI - Empirical evidence and optimal policy”, of December 3, 2025, PE 778.859, available at https://www.europarl.europa.eu/thinktank/en/document/IUST_STU(2025)778859, last accessed at March 20, 2026.
[62] See page 16 of the Study.
[63] See the Jun Ito, on the outcome of the DABUS case in Japan, the Decision of the Tokyo District Court of May 16, 2024, Reiwa 5 (Gyo-U) 5001, in GRUR Patent. Patentrecht in der Praxis, no. 7/2024, pp. 307-308; Senftleben, Martin and Buijtelaar, Laurens, Robot Creativity: An Incentive-Based Neighboring Rights Approach (October 1, 2020). Available at SSRN: https://ssrn.com/abstract=3707741 or http://dx.doi.org/10.2139/ssrn.3707741, last accessed on March 16, 2026.
[64] Senftleben, Martin and Buijtelaar, Laurens, ”Robot Creativity: An Incentive-Based Neighboring Rights Approach” (October 1, 2020). Available at SSRN: https://ssrn.com/abstract=3707741 or http://dx.doi.org/10.2139/ssrn.3707741, last accessed March 21, 2026.
[65] France Vehar, Thomas Gils are of opinion that AI can create ”works” even completely without human intervention, op. cit. supra, p. 721.
[66] Nicola Lucchi, op. cit., page 8 and 11.
[67] Nicola Lucchi points that the current legal model ”highlights the growing difficulty in assessing the threshold of human involvement in AI-assisted creativity”, op. cit., pages 94-98. For doctrinal discussions in Germany, see France Vehar, Thomas Gils, op. cit, p. 721-722. See also Hilty, Reto and Hoffmann, Jörg and Scheuerer, Stefan, ”Intellectual Property Justification for Artificial Intelligence”, op. cit.
[68] Nicola Lucchi, op. cit, p. 23.
[69] Senftleben, Martin and Buijtelaar, Laurens, op. cit.
[70] Technical Aspects of Artificial Intelligence [...], op. cit.
[71] Available at https://www.copyright.gov/ai/, last accessed March 20, 2026, page 6.
[72] Alexis Léautier, Panorama et perspectives pour les solutions de détection de contenus artificiels, available at https://linc.cnil.fr/panorama-et-perspectives-pour-les-solutions-de-detection-de-contenus-artificiels-12, last accessed March 20, 2026, cited by Benhamou, Yaniv; Andrijevic, Ana, ”Intelligence artificielle générative d'images et droit d'auteur [...], op. cit.
[73] Nicola Lucchi, op. cit, p. 97.
[74] Péter Mezei, op. cit.
[75] Available at https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32019L0790, last accessed on March 17, 2026.
[76] Hilty, Reto and Hoffmann, Jörg and Scheuerer, Stefan, ”Intellectual Property Justification for Artificial Intelligence” [...], op. cit.
[77] Iaia, Vincenzo, op. cit.; Péter Mezei, op. cit. s
[78] Christian Peukert, The Study for the EP JURI Committee, op. cit., p. 17.
[79] Martin Senftleben, ”Text and Data Mining, Generative AI, and the Copyright Three-Step Test”, IIC (2026) 57:67–107, https://doi.org/10.1007/s40319-026-01680-2.
[80] Drexl, Josef and Hilty, Reto and Desaunettes-Barbero, Luc and Globocnik, Jure and Gonzalez Otero, Begoña and Hoffmann, Jörg and Kim, Daria and Kulhari, Shraddha and Richter, Heiko and Scheuerer, Stefan and Slowinski, Peter R. and Wiedemann, Klaus, ”Artificial Intelligence and Intellectual Property Law - Position Statement of the Max Planck Institute for Innovation and Competition of 9 April 2021 on the Current Debate” (April 9, 2021). Max Planck Institute for Innovation & Competition Research Paper No. 21-10, Available at SSRN: https://ssrn.com/abstract=3822924, last accessed March 21, 2026; Peter Georg Picht, Florent Thouvenin, op. cit.
[81] Available at https://www.gov.uk/government/consultations/artificial-intelligence-and-ip-copyright-and-patents/artificial-intelligence-and-intellectual-property-copyright-and-patents, last accessed on March 17, 2026.
[82] Of June 28, 2022, available at https://www.gov.uk/government/consultations/artificial-intelligence-and-ip-copyright-and-patents/outcome/artificial-intelligence-and-intellectual-property-copyright-and-patents-government-response-to-consultation, lat accessed March 17, 2026.
[83] The US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 36.
[84] The European Commission Final report, “Trends and Developments in Artificial Intelligence – Challenges to the Intellectual Property Rights Framework”, of September 2020, available at https://op.europa.eu/en/publication-detail/-/publication/394345a1-2ecf-11eb-b27b-01aa75ed71a1/language-en, last accessed on March 20, 2026.
[85] 2025/2058 (INI) P10_TA (2026) 0066, available at https://www.europarl.europa.eu/doceo/document/TA-10-2026-0066_EN.html, last accessed March 20, 2026.
[86] Peter Georg Picht, Florent Thouvenin, op cit.
[87] The US Copyright Office Report on Copyright and Artificial Intelligence, Part 2, page 31; France Vehar and Thomas Gils, op. cit, p. 726.