Introduction
Welcome to the third quarterly Copyright and Generative AI (GAI) blog post of 2026. If you’ve come across this page, you’re likely working in the copyright field in some capacity. As such, the current debate surrounding what, if any, changes should be made to copyright law to account for advancements in GAI has likely piqued your interest. With so much information and so many opinions swirling around the internet from all sides of the debate, it can be hard to keep track of where the conversation stands. We hope these blogs will help make sense of it all.
Every three months, you can expect an updated post that includes important cases to watch, recent international developments, a summary of the latest discussions (with an effort to remain impartial) and finally, a suggested long-form article to help you dive deeper into one aspect of the subject.
So, without further ado, let’s get into it!
Cases to Watch
Before we get into the specific cases we will be highlighting, it’s important to note that the number of cases between copyright holders and generative AI companies in the United States alone reached 125 in August and is now over 145. If nothing else, this shows how contentious the issue remains even as generative AI continues to develop at an unprecedented pace. As such, we’ve decided to highlight two current US cases in this section as well as ones in India and Germany.
One of the more recent cases in the US saw nearly 400 local newspapers sue OpenAI and Microsoft, alleging copyright theft. By arguing that the two tech companies unfairly made use of innumerable pieces of local journalism to train their models, the publications insist that this could spell disaster for their future existence. Essentially, if users turn to models such as ChatGPT for information on local news, their already threatened publications will lose the crucial readership that allows them to continue paying journalists to report on local news. Though lost ad revenue is not necessarily a valid argument in copyright cases, the possibility of the plaintiffs claiming market displacement might strengthen their case that fair use should not apply in this instance.
Smaller newspapers are not the only ones claiming market harm due to the actions of the makers of GAI models. Back in 2023, the New York Times (NYT) became the first major publication to sue OpenAI and Microsoft for copyright infringement. This case is still ongoing but relies on the same facts, namely that tools such as ChatGPT allow users to bypass NYT paywalls and access its content simply by prompting the GAI models to summarize reporting. The NYT complaint further alleges that results from the GAI model can be hallucinated and therefore mislead users to believe that the NYT has reported on, or confirmed, certain unsubstantiated facts (for example, attributing falsified quotes to prominent individuals).
Though the case is ongoing, the NYT is now seeking to amend its complaint due to the recent Cox Communications decision, which made it imperative for plaintiffs to prove deliberate contributory infringement by the companies that created the tools being used. As a reminder, the decision in Cox Communications found that the internet provider could not be found liable for contributory infringement unless it clearly intended its service to be used for infringing purposes. The NYT is therefore seeking to amend its complaint to further allege that Microsoft specifically created a custom supercomputer to allow OpenAI to target the NYT’s content for ingesting into its GAI models. While the NYT has agreed to drop two complaints in return for obtaining this amendment, the defendants continue to insist that their use of the plaintiff’s content constitutes fair use. However, it recently came to light that OpenAI may have been deceptive about its ability to produce de-identified ChatGPT logs, which may provide evidence of market harm and infringement (evidence they have been fighting to keep out of the courts since the trial began). The NYT is now calling for OpenAI to face sanctions for hiding crucial evidence, which could turn the tide in this case. If the amendment and sanctions succeed, the most drastic result could be that OpenAI will have to start from scratch.
In the most recent update on this case, the US government recently filed a Statement of Interest in support of OpenAI’s fair use argument which, while not binding, could potentially strengthen the defendant’s argument while simultaneously augmenting the number of arguments the plaintiff has to answer. We’ll definitely be keeping our eye on this case.
ANI Media Pvt Ltd v. Open AI OPCO LLC, CS(COMM) 1028/2024 (Delhi High Court, 19 November 2024) INDIA
OpenAI recently won the first round of a copyright-infringement trial in India when the High Court of Delhi rejected Asian News International’s (ANI) application for a temporary injunction against OpenAI. While this in no way concludes the case overall, it is important to note that this rejection included a thorough overview of both the input and output stages of the GAI process as well as the Indian fairness test to determine whether the model’s use of copyright-protected works can be considered fair. So, though this step is not conclusive, it gives an idea of where the judge might be leaning (without further evidence being presented), which seems to be in favour of the defendants. Fascinatingly, the judge relied on several US cases to come to this conclusion, including the Kadrey decision and the much older Google Books ruling. Another crucial case to keep an eye on!
GEMA v. Suno Inc. case no. 42 O 763/25 (Munich District Court, 21 January 2025) GERMANY
In the music realm, we recently saw yet another landmark decision in Germany in a case between the German collective society GEMA and the GAI company Suno. Focused on six musical works (including “Mambo No. 5”), GEMA accused Suno of reproducing these works during the training process and storing them through memorization. Germany’s Munich Regional Court eventually found in favour of the plaintiffs on the issue of copyright infringement. Though this seems to be the complete opposite tack to fair use decisions in the US, not to mention the exact opposite of the US government’s recent Statement of Interest, the Munich court did devote 10 pages to an analysis of US fair use and jurisprudence and therefore provided some helpful information about when a use can be considered transformative or not.
There are many other cases we could highlight, all of which will play a crucial role in determining the relationship between IP and GAI moving forward. You can see an excellent overview of the cases currently before the courts globally here. It will be interesting to see if any of the cases mentioned above make it into our next roundup in one form or another.
International Developments
In international developments, there has been a lot of movement on the question of transparency and GAI as well as questions about ownership of outputs. Canada, in particular, just closed its consultation on AI and transparency, which included questions about labeling of AI output. While we will need to wait a while until we have access to the results of this consultation, the EU recently published its own guideline on AI labeling as a response to transparency concerns. Although predictably criticized by GAI companies, the guideline is already seeing results with many of these same companies turning to watermarks (whether easily visible or not) as a tool to label content being produced using GAI. However, the key point is that Article 50 of the EU AI Act requires such content to be clearly labeled regardless of where the model is based. As a result, Anthropic recently announced that it will now embed watermarks directly in text produced using Claude. Soon afterwards, a developer released a tool that he said can remove the watermark produced by Claude (though he insists his motives had nothing to do with copyright concerns). With the possibility of such tools, the question remains whether GAI companies will be found to be compliant if they embed watermarks despite the potential for those watermarks to be removed. This is just the beginning of implementation, though large tech companies such as Google and Meta are already making changes. It will be interesting to see how GAI companies interpret these new regulations, especially considering the large fines they face if they don’t comply.
Regarding ownership in works produced using GAI, Saudi Arabia decided to dodge this question entirely in its latest update to the country’s copyright laws. Though these new regulations include the addition of a sweeping exception for GAI training, noticeably, they do not tackle the issues of copyright applicability to works produced using GAI or authorship requirements. In India, on the other hand, the country’s Copyright Office recently refused Dr. Stephen Thaler’s attempt to register the copyright in his A Recent Entrance to Paradise work citing the GAI model DABUS as the author. Interestingly, the Office concluded that the GAI-produced work did indeed meet the originality requirements (which might tie into other discussions around infringement) but that the refusal was based on the fact that a machine cannot be listed as the author of a copyright-protected work.
Another ownership question being grappled with is that of AI twins and who owns the output they produce. Essentially, an AI twin is a closed GAI model trained exclusively on one person’s life’s work (including, but not limited to, research, presentations and publications), allowing the model to access the knowledge of the person whose works it has ingested. An important aspect of this conversation is that these AI knowledge twins are not only being used by the people on which they are based but are also sometimes created for public use. While this can help knowledge leaders reach even larger audiences without adding a burdensome time commitment, the question remains: Who owns the copyright in content produced using these models? The Kluwer Copyright Blog has written a fascinating article on this topic, positing that there are three possible answers to this question: the owner of the AI twin, the user of the AI twin or, well, no one (i.e., it would be in the public domain). I’m not sure what the answer is, but I am sure this is not the last we’ll hear about this conversation!
Summary of Latest Discussions
First and foremost, we have seen a rising trend of enormous book orders being placed with small and large booksellers from mysterious origins (sometimes thousands of books in a single order). While the income is obviously welcome to many booksellers whose businesses have struggled in recent years, some are concerned that these orders are being placed by companies supplying GAI models with physical books to train on, which are then destroyed during that process. While the destruction of books (often called pulping in industry-speak) is common practice, there is some concern that out-of-print or rarer editions are being swept up in this destructive form of GAI training. This conversation touches on copyright, as the scanning of the books for training may constitute an infringing reproduction (depending on where you stand in this debate); however, it also raises the larger question of the value of human creativity and whether we as a global community are willing to see some of it possibly destroyed in the name of technological development.
As a potential example to other countries (aside from the EU, which has taken large strides forward in regulation), Australian Prime Minister Anthony Albanese recently announced that his Parliamentary Office will now be taking direct control of the country’s AI agenda. Though the Australian government recently rejected the possibility of implementing a text and data mining (TDM) exception for GAI, it is being solicited by tech companies with hard-to-refuse deals in exchange for clarifying the copyright question. For example, Anthropic has proposed a US$15 billion investment in Australian AI infrastructure, which relies entirely on the complexities between copyright and GAI being settled. With such a large potential investment in the country on the table, some are questioning whether the decision to take control of the AI agenda moving forward is indicative of the will to make some big changes to national law soon. For the time being, the Australian government seems to be considering several different types of licensing regimes, including the oft-criticized system of compulsory licensing, but no decisions have yet been taken. Yet another potential development to keep a close watch on.
Deep Dive
For this month’s deep dive, we are suggesting the UK House of Lords’ Report on AI, copyright and the creative industries. Though this report was published by the Communications and Digital Committee back in March, it is worth a read as it goes into detail about the various complications involved in balancing the needs of both creatives and the tech sector. Have a read and let us know what you think!
More Information
That’s it for this quarter’s update. However, if you’re looking for more regular and up-to-date news about the nexus of copyright and AI, check out our free newsletter, Copyrighting AI. And if you have any thoughts on what we have shared here, please feel free to reach out!
