At the recent User Rights conference convened in Geneva by the Centre on Knowledge Governance, one panel addressed the topic of the Globalization of Fair Use. As the moderator of the panel, I observed that in a five-week period this past summer, three different courts on three different continents addressed the U.S. fair use doctrine, including references to decisions from U.S. courts as well as scholarship concerning the fair use doctrine. (Other conference panels touched on these decisions as well.) These decisions demonstrate the increasing familiarity of courts around the world with fair use and growing comfort with its flexible, open-ended approach. More generally, the opinions show the globalization of legal research and judicial reasoning; courts in South Africa, India, and Germany all have ready access to U.S. legal resources, and feel competent to rely upon them.
I. South Africa Constitutional Court
On June 26, 2026, the South African Constitutional Court issued its judgment in In re Constitutionality of the Copyright Amendment Bill and the Performers’ Protection Amendment Bill. The Constitutional Court’s decision is the most recent development in this legislation’s two-decade journey. In 2020, the South African Parliament adopted the Copyright Amendment Bill (CAB) and the Performers’ Protection Amendment Bill. The President of South Africa referred the Bills back to Parliament after identifying concerns with the constitutionality of some of their provisions, most notably the CAB’s exception for fair use. Parliament resubmitted the Bills for the President’s signature in 2024 after making some amendments—but none to the fair use provision. Soon thereafter, the President referred the Bills to the Constitutional Court, questioning the constitutionality of some of their provisions. In particular, the President queried whether the fair use provision in section 12A of the CAB contravened South Africa’s international treaty obligations and therefore was an unconstitutionally arbitrary deprivation of property.
In its June 2026 decision, the court ruled that the CAB’s fair use provision was constitutional. The court noted that criticism raised by some parties that the CAB replaced the existing fair dealing provision based on the English Copyright Act of 1911 with a more flexible and open U.S.-style fair use provision. The court observed in ¶ 114 that
the strict distinction between “fair dealing” and “fair use” is no longer appropriate. A number of countries have implemented schemes that, while being labelled fair use, contain relatively clear categories of permitted use, having combined these schemes with elements more traditionally associated with fair dealing. On the other hand, Canada, on paper, has a traditional fair dealing regime, which in practice is interpreted widely, making it operate more akin to a fair use regime. It must therefore be emphasised that fair use and fair dealing are not to be seen as completely distinct simply because of their terminology, but instead may “lie on two ends of a continuum, with both requiring the case-by-case balancing of multiple fairness factors”. The implementation of fair use regimes has not produced the expansive, unpredictable outcomes feared by US based copyright industries; rather, these regimes have often been implemented in a narrower and more restrained manner than the US jurisprudence itself.
In support of the proposition of the inappropriateness of a strict distinction between fair use and fair dealing, the court cited Peter Yu’s article “Customizing Fair Use Transplants.” In support of the proposition that the non-U.S. fair use regimes have been implemented more narrowly than the U.S. jurisprudence, the court cited Niva Elkin Koren and Neil Netanel’s article “Transplanting Fair Use across the Globe: A Case Study Testing the Credibility of U.S. Opposition.”
In ¶ 115, the court explained that CAB section 12A “substantially mirrors” the fair use provision in section 107 of the U.S. Copyright Act: “This includes the provision for an open list of purposes of use and a list of particular factors to be considered in assessing whether the use is fair….” The court then identifies salutary features of the open fair use approach:
The shift from fair dealing to fair use denotes an open and adaptable approach to copyright exceptions, one better able to evolve alongside creative and cultural developments. Furthermore, the open list of purposes is preferable to the more restricted and closed list of purposes in subsection 12(1) of the Act. It mirrors the same underlying rationale of the present purposes, that of public benefit, and it appears to be of equal importance as those present in subsections 12(1)(a)-(c) of the current Act, those purposes being criticism and review, reporting on current events and research and private study. Rather than limiting use to a closed list of permitted purposes, the provision is best understood as a mechanism for determining whether a user’s conduct is fair, rather than as a set of requirements to be met. Ultimately, the facts will indicate whether the conduct is fair.
The court observed that several other countries have incorporated fair use provisions into their law, including Taiwan and the Philippines. The court concluded that “while the lack of challenges in the international context against fair use provisions is not decisive of their conformity to international law, it serves as an indication that, in allowing limited fair use for specified purposes that are part of a statutory framework, such fair use provisions pass the test in Article 9(2) of the Berne Convention.” ¶ 118.
The court then discussed whether the fair use provision constituted an arbitrary deprivation of property. While the provision might lead to the deprivation of property, this deprivation is not arbitrary. The court rejected the argument that the provision was fatally vague or that the inclusion of the phrase “such as” unduly widened the scope of works that could be used freely:
If one examines the various components of the provision, it is reasonably possible to ascertain its meaning and what it expects of users, or the burden it imposes on them. In subsection (a) one can clearly see in which instances copyright infringement will not apply and for what purposes. The phrase “such as” implies that there are other purposes not mentioned that would fall into the exception. To interpret this as “any other potential use”, however, would be an absurd interpretation of section 12A if regard is had to the purpose of the express exceptions. The purpose of the express exceptions is to allow for the use of copyrighted work for important public purposes or for individuals to use works in their personal capacity. Therefore, the phrase “such as” can only have the effect of including other purposes not yet contemplated by the Legislature but which share attributes similar to the named purposes, that is, public benefit or private and personal use. ¶ 132.
Subsection 12A(b) goes further by providing guidance on which factors will be taken into account in determining when the fair use exceptions will apply. Not only does the provision provide a range of express purposes for fair use, but it also sets out a framework of the factors to be taken into account in determining such fair use. Again, the phrases “including but not limited to” and “including” do not leave the door entirely open to any factors. The factors that are “included” but do not form part of the list are not random. These must be factors that align with the purpose of the exception. As this Court emphasised in Affordable Medicines and Savoi, generality does not mean that a provision is vague. In fact, generality may be necessary to provide for flexibility in the event that persons are confronted with circumstances that do not neatly fall within the express provisions. This is precisely the function of the phrases “such as”, “including” or “included, but not limited to”, namely, to provide flexibility for exemptions which the Legislature has not yet contemplated. ¶ 133.
Because section 12A’s potential deprivation of property was not arbitrary, it was not unconstitutional. Although the court also found most of the other challenged provisions to be constitutional, it did find a single exception relating to education to be an unconstitutionally arbitrary deprivation of property.
II. Delhi High Court
On July 24, 2026, the Delhi High Court issued a preliminary judgment in ANI Media v. OpenAI, finding that OpenAI’s ingestion of ANI Media’s works in the course of training its large language model was a fair dealing under section 52(1)(a) of the Indian Copyright Act. In the course of its analysis, the court repeatedly relied on foreign precedents, including the U.S. fair use decision in Bartz v. Anthropic.
At the outset of the opinion, the Delhi court noted
This case is a perfect illustration of how the law must continually catch up with technology. The case has to be adjudicated on the anvil of the Copyright Act, originally enacted in 1957, though amended from time to time. Clearly, when the Copyright Act or any of its amendments were enacted, the legislature could not have imagined the vast technological advances in the field of AI. ¶ 7.
The Delhi court observed that
in these uncharted areas, Courts often take aid from judgments of foreign Courts on similar issues. On various occasions, while dealing with Intellectual Property laws, the Indian Courts have recognized the importance of global doctrinal interdependence. The Courts laid emphasis on harmonization of basic principles of laws in foreign jurisprudence with the national laws, especially when the national laws do not indicate a contrary intent. However, while doing so, the Courts must be mindful that the underlying legislation in foreign jurisdictions may be quite different from that prevailing in India. ¶ 8.
When the Delhi court framed its understanding of fair dealing under section 52(1)(a), it considered how similar exceptions have been interpreted internationally. It noted that the Supreme Court of Canada in CCH Canadian v. Law Society held that the “fair dealing exception is an integral part of the Copyright Act (of Canada) rather than simply a defence. Therefore, in order to maintain a proper balance between the rights of copyright owners and users, it should not be interpreted restrictively.” ¶ 165. Accordingly, the Delhi court concluded that section 52 of the Indian Copyright Act “strikes a balance between the exclusive rights granted to the owner of the copyright and the competing public interest of encouraging creativity and disseminating knowledge. Hence, Section 52 is not in the nature of a proviso or an exception to the Section 51 of the Copyright Act.” ¶ 166. Rather, in the Delhi court’s view, “Section 52 independently defines the rights/privileges of a user in respect of copyrighted works and must not be interpreted in a restrictive manner. It is an integral part of the Copyright Act and therefore, it has to be given broad and liberal interpretation.” ¶ 167.
The court first considered the statute’s requirement that the copy accessed by the defendant be a non-infringing copy. Here, OpenAI accessed content that was freely available on ANI’s website. The court contrasted this with Bartz v. Anthropic, where the district court “clearly made a distinction between acquiring plaintiff’s books in a lawful manner and obtaining copies from ‘Shadow Libraries.’” ¶ 199. The Delhi court observed that “the US Court held that scanning of lawfully acquired books amounted to fair use, whereas obtaining copies from shadow libraries was not fair use.” Id.
The Delhi court next considered whether the ingestion for training was a permitted purpose under the Act. The court specifically focused on whether this was a “private or personal use, including research.” Relying on the Merriam Webster and Cambridge Dictionaries, as well as Indian precedents, the court concluded that the term “research” should be
given an updating construction by taking into account the modern-day technical advancements. With the advent of these technologies, research/learning is no longer confined to humans. It is now being done through Artificial Intelligence. However, ultimately the research is at the behest of humans and for the benefit of humans.[…] [T]he acts of further research cannot be confined to acts of human beings alone and the same would extend to machine learning as well. ¶ 218.
The Delhi court next turned to the fairness of OpenAI’s research acts under the fair dealing statute. After considering Indian precedents, the court concluded that there was no single test for “fairness” adopted uniformly and consistently by Indian courts, and that the U.S. four factor test was not applicable in India. The court, therefore, developed its own factors for determining fairness, which it asserted were “broadly in line with the principles emerging from” the Berne Three Step Test:
a. Whether OpenAI’s use of ANI’s original literary works is limited to training its LLMs underlying ChatGPT?
b. Whether the usage of ANI’s literary works by Open AI would result in economic competition and would prejudice the legitimate interests of ANI, thereby causing actual or potential damage to ANI?
c. Whether the functions performed by OpenAI through ChatGPT serve the overall public interest?
The court proceeded to weigh these factors. With respect to the first factor, the court found that the record established that all copying by OpenAI was for the purpose of training its AI.
With respect to the second factor, the Delhi court relied on the finding in Bartz v. Anthropic that the use of plaintiffs’ works “to train specific LLMs did not and will not displace demand for copies of Authors’ works, or not in the way that counts under the Copyright Act.” The Delhi court also noted that the Bartz court found that “using copyrighted works to train LLMs to generate new text was quintessentially transformative.” Further, the Delhi court cited the U.S. district court decision in Kadrey v. Meta holding that Meta’s purpose in training an LLM was likewise transformative. The Delhi court also cited the decision of the U.S. Court of Appeals for the Second Circuit in Authors Guild v. Google holding that Google’s display of snippets was highly transformative and not a form of market substitution. Based on these U.S. precedents and its review of the facts in the record, the Delhi court ruled that OpenAI’s use of ANI’s works for AI training would not result in market substitution not any other actual or potential damage.
Turning to the third factor, whether the functions performed by OpenAI through ChatGPT would serve the public interest, the court stated that training the LLM that underlies ChaptGPT
contributes to advancing scientific knowledge, developing innovative computational tools, disseminating information, promoting education, enhancing accessibility, and advancing artificial intelligence research. These objectives resonate with the broader constitutional and statutory commitment to advancing knowledge and learning. While the rights of copyright owners remain important, the societal benefits arising from scientific and technological research constitute a relevant consideration in assessing the fairness of a dealing. ¶ 253.
III. Munich Regional Court
On July 31, 2026, a German court in GEMA v. Suno rejected the application of fair use to AI training that occurred in the United States. The court reached this determination by distinguishing, not disparaging, the Anthropic and Kadrey decisions.
GEMA is a German collecting society for musical works. It alleged that Suno trained it AI model on, inter alia, six musical works administered by GEMA. GEMA claimed although the training occurred in the United States, these works were memorized in the model, that the model was stored on German territory, and that the model regurgitated the works in Germany. On July 31, 2026, the Munich Regional Court found that Suno infringed German copyright law with respect to the activities the occurred in Germany. In particular, the court found that Suno’s activities did not fall within Germany’s implementation of the text and data mining exception in the EU Directive on Copyright in the Digital Single Market.
Furthermore, the German Collecting Societies Act permits a collecting society to bring all claims against an infringer in a German court that has jurisdiction over the infringer for at least one of those claims. Thus, because the court had jurisdiction over GEMA’s claims against Suno arising in Germany, the court was also able to exercise jurisdiction over GEMA’s claims against Suno arising out of Suno’s activities in the United States. However, the court had to evaluate these activities under U.S. law. Accordingly, the Munich court considered the permissibility of Suno’s copying in the course of training in the U.S. under fair use. The court distinguished Bartz v. Anthropic and Kadrey v. Meta on the factual grounds that Suno’s model memorized and regurgitated the works at issue, while the plaintiffs proved no memorization and regurgitation in Bartz and Kadrey. The Munich court quote quoted the Bartz court’s statement that “if the outputs seen by users had been infringing, Authors would have a different case.”
Suno has appealed the Munich court’s decision, arguing that its model did not memorize; but rather, that GEMA engineered substantially similar outputs through its excessive number of prompts. However, if memorization did occur, it is hard to fault the Munich court’s interpretation of the fair use provision.
IV. Conclusion
Rightsholders have long portrayed the U.S. fair use doctrine as an international copyright anomaly, even suggesting that it did not comply with the Berne three-step test. They further have argued that civil law judges are incapable of applying the fair use factors, which are rooted in common law. These three decisions from the summer of 2026 belie these assertions. The decisions from South Africa and India undermine the claimed distinction between fair use and fair dealing, thereby placing a third of the world’s population in a jurisdiction recognizing an open, flexible copyright exception. Both of those decisions also demonstrate the compatibility of fair use with the Berne three-step test. Finally, the GEMA decision shows that civil law judges are able to comprehend and apply the fair use factors with their case-by-case common law approach.
