As we enter a new year, we would like to take this opportunity to pass on our best wishes for 2025 to all of our readers, as well as reflect on developments in copyright over the past year. Last year was another busy one in the copyright world, with an increasing focus on the relationship between AI and copyright, ongoing European copyright reform, a number of landmark CJEU decisions and notable developments in a number of jurisdictions.
Here is a quick look back at our 10 most-read posts last year:
1. Is Generative AI Fair Use of Copyright Works? NYT v. OpenAI by Mira T. Sundara Rajan
The case recently brought against OpenAI by the New York Times is the latest in a series of legal actions involving AI in the United States, and mirrored in other countries –notably, the UK. In order to train their technologies, should AI companies be allowed to use works under copyright protection without consent? The lawsuits brought by the owners of such works, including artworks in the case of image-generators and journalism in the NYT case, claim that this should not be allowed. Such uses, they argue, constitute copyright infringement.
2. Beijing Internet Court Grants Copyright to AI-Generated Image for the First Time by
On November 27, 2023, the Beijing Internet Court (BIC) ruled in an infringement lawsuit (Li v. Liu) that an AI-generated image is copyrightable and that a person who prompted the AI-generated image is entitled to the right of authorship under Chinese Copyright Law. Plaintiff generated an image of a woman by using Stable Diffusion, an open-source generative AI model that creates images from textual prompts. After publishing the disputed image on a Chinese social media platform (Xiaohongshu), Plaintiff discovered that Defendant had used the same image to illustrate an article on a different website without permission. Plaintiff then sued Defendant in the BIC.
3. EU law: Generative AI, copyright infringements and liability – My guess for a hot topic in 2024 by Jan Bernd Nordemann
While some digital topics have been known to cause a great commotion in copyright circles only to later sink practically without a trace, unless I am mistaken, the issue of the copyright implications of AI is different. One AI topic, which has so far only been examined in any depth in relation to EU copyright law in a few instances, is copyright infringement by generative AI and the associated liability. In this regard, there are two aspects that need to be looked at separately, namely when does AI output constitute an infringement and who is liable for copyright-infringing AI output?
4. Machine readable or not? – notes on the hearing in LAION e.v. vs Kneschke by Paul Keller
[T]he District Court of Hamburg, Germany, held a hearing in the first European case to examine the legality of using copyrighted works for the purpose of training generative AI models.
The case centers on LAION e.V.’s (a German non-profit organization that builds widely used training datasets) download of an image by German photographer Robert Kneschke for inclusion in the LAION 5B dataset. Neither party disputes that the image in question was downloaded, analyzed, and subsequently included in the training dataset, but LAION claims that this is legally permissible, while Kneschke disputes this. The disputed image was freely available without a paywall on the website bigstock.com.
5. Are AI models’ weights protected databases? by Nuno Sousa e Silva
The ongoing Artificial Intelligence (AI) revolution has machine learning models at its core. Contrary to classic computer programs written by developers, many of these models rely on vast artificial neural networks trained in giant amounts of data. […] The final result, i.e. the model, consists of two files – a simple run file that establishes the functioning of the model (the model architecture) and a much larger file of parameters or weights (expressed as floating point numbers). Weights are a mathematical expression of the connection between the neurons that make up the network. […] Much has been written about the legal challenges and qualifications of the training process (knowing whether it’s legally permissible to train these models on copyright-protected material) and the outputs of these models (especially if there is copyright in the results generated)… However, it’s unclear whether the models themselves are currently protected by intellectual property laws. The run file is a classic piece of software and does not pose particular difficulties. It’s the parameters or weights (the larger file) that raise the puzzling questions.
6. Meta’s AI Arriving in Europe: Privacy Disputes Concealing Copyright Concerns by
Since 22 May 2024, Meta has notified to European users of Instagram and Facebook – through in-app notifications and emails – an update of its privacy policy, linked to the upcoming implementation of artificial intelligence (AI) technologies in the area.
Indeed, the company already developed and made available some AI features and experiences in other parts of the world, including an assistant called “Meta AI”, built on a large language model (LLM) called “Llama”, and, in an official statement, announced the imminent plan to expand their use also in Europe.
This initiative resulted in some pending privacy disputes, which polarized the debate. However, data appear to be just one side of the medal and to conceal much deeper copyright concerns. Given the holistic approach required by the challenges related to the construction of AI models, it is appropriate to proceed in order, starting with a broader overview.
7. TDM: Poland challenges the rule of EU copyright law by Paul Keller
When life gives you lemons, make lemonade. This must have been the key insight at the Polish Culture and National Heritage Ministry when the new administration took over and discovered that more than 2.5 years after the implementation deadline, Poland still had to implement the provisions of the 2019 Copyright in the Digital Single Market Directive into national law. So how do you make lemonade out of the fact that you are the only EU Member State without an implementation? You claim that the delay allows you to propose a better implementation.
8. UK’s short-lived dream for a code of practice on genAI and copyright law by Alina Trapova
The UK’s attempt to deal with generative AI, training data and copyright law has taken yet another turn. On 6 February 2024, in its response to the AI White Paper consultation, the UK government announced that it will drop its plans for a code of practice on copyright and AI – a work it has been carrying out for less than a year since its announcement in May 2023.
9. Memorisation in generative models and EU copyright law: an interdisciplinary view by
Large language models’ (LLMs) greatest strength may also be their greatest weakness: their learning is so advanced that sometimes, just like humans, they memorise. This is not surprising, of course, because computers are really good at mainly two things: storing and analysing data. There is now empirical evidence that deep learning models are prone to memorising (i.e., storing) fragments of their training data. Just like the human brain needs to memorise fragments of information to learn, so do LLMs. And when they reproduce verbatim these fragments, this may be a ground for copyright infringement.
10. Open Source AI – definition and selected legal challenges by Yaniv Benhamou
In the generative AI era, there is a proliferation of open source claims (i.e. operators that claim to release AI models sufficiently open to be part of the open source or open innovation movement, as opposed to closed-source model), such as open source and open access foundation models (e.g. Google BERT, Meta LLaMA Large Language Model (LLM), OpenAI API). While an open source approach to AI is valued as important for fostering innovation and competition, the notion raises many questions: (1) What is open source AI? Which elements shall be available as open source? Can it be everything (i.e. all elements composing the AI model) or only specific components (e.g. training data, weighting factors)? (2) What is the intersection and the difference between ‘open data’ and ‘open source’? (3) What is the effect of open source licenses on the AI model that uses only some open source components? (4) What is the liability of open source contributors? (5) What is the impact of new regulation on open source AI?
