Professor Eduardo Alonso, School of Science & Technology, and Dr Enrico Bonadio, The City Law School, host an international workshop to discuss the latest developments in the field.

By Verónica Muñoz Martínez, Press Office Gradvantage Intern

On Monday 22 July, The Artificial Intelligence Research Centre (CitAI), based at City, University of London, hosted a hybrid event that delved into the intersection of artificial intelligence (AI) technology and its legal, ethical, and social implications.

The event featured both AI and legal experts who discussed various aspects of AI, particularly focusing on copyright infringement and the potential remedies.

The workshops was organised by Dr Enrico Bonadio, Reader in Intellectual Property Law at The City Law School and Professor Eduardo Alonso, School of Science & Technology, under the auspices of City, University of London’s Artificial Intelligence Research Centre.

Twenty-three experts from around the world took part, from institutions in Austria, France, Germany, Hungary, Israel, Italy, Portugal, The Netherlands, Norway, Spain, Turkey and the UK.

Each participant delivered a talk on their area of expertise, and many attended the workshop in person.

Professor Eduardo Alonso providing a technical introduction to generative AI

Professor Alonso provided the opening remarks to the workshop and provided a technical introduction to what ‘generative’ artificial intelligence is, saying in part:

From the 60s and 70s, we started with symbolic artificial intelligence. Basically, you have some premises between a formal, logical language and with that you manipulate those premises with some well-established rules up to some kind of conclusion we call a theory.

This was not particularly effective, it’s what we call now non-generative artificial. However, even within this paradigm, people started to do things that are even more creative. For example, and I like this picture — because it gives us a sense of a system of the nodes as the segments, and the links between the nodes are the rules that link one to another, but as you see in the nodes, you have a die, that represents that you are introducing a stochasticity [randomness] that the system becomes more creative, because, allegedly, you can’t control how these non-deterministic rules will affect the conclusions.

Migle Laukyte, Pompeu Fabra Barcelona, delivering her talk - Science, Knowledge and IP: the case of AI

Antoni Rubi Puig is Associate Professor in Civil Law at the Universitat Pompeu Fabra, Spain.  He presented his insights on AI model ‘disgorgement’ as a remedy for copyright infringement of the model. He highlighted the significant environmental impact and the holdup problems caused by removing an entire AI model and rebuilding it.

Rubi Puig argued that if the model was licensed to third parties that have developed application systems based on the initial model, retraining might affect these applications in the downstream chain of value.  He proposed a ‘machine-selective’ forgetting or unlearning approach as a more feasible solution.

Instead of deleting the infringing or local data and retraining the model, machine learning removes the effects of such data on the model," he explained. He is positive about this approach, adding: “Machine unlearning is a novel area in AI, and we can make several improvements in the years ahead.

He also cited Article 17 of the EU Artificial Intelligence Act as a legal remedy for AI infringement. Discussing the Act, Rubi Puig examined the legal framework for Intellectual Property Rights (IPR) enforcement in the EU, sharing:

According to Article Three of the directive, we find a balance test or a trade-off. This forces us to think about measures, procedures, and remedies that are fair and equitable, and shall not be unnecessarily complicated or costly.

Rubi Puig explained that payment of damages could be an alternative to model disgorgement. “We have this option of providing for the possibility, in appropriate cases, of pecuniary compensation as an alternative measure for corrective measures,” he said. He emphasized the need to consider the impact of these remedies on third parties, especially those in the generative AI value chain.

Joao Pedro Quintais, an Associate Professor of Law from the University of Amsterdam, focused on the AI Act and Copyright Acquis in his presentation.

Discussing copyright infringement in generative AI, he made a clear distinction between AI systems, GPI systems (ChatGPT, Dall-E) and GPAI System (GPTs and non-gen). He said:

If you develop an AI model you are working on TDM. GPAI model providers must create plus publicly share.

TDM applies mostly to the training stage of developing an AI model, and there are copyright issues with the model itself and the way it's incorporated into the system. He explained that large generative AI models are subject to copyright provisions, but the obligations only apply to providers who are vertically integrated within the GPI system. Dr Bonadio, shared insights from his research on AI’s impact on street art and graffiti. In his study "Copyright in the Street: An Oral History of Creative Processes in Street Art and Graffiti Subcultures" he interviewed more than 600 graffiti artists who had different opinions on AI-generated graffiti. His presentation touched on the potential of AI to enhance street art, exploring creative possibilities, censorship concerns, and artist’s permission.

“Some artists feared their work being copied without consent, while others were unimpressed by AI-generated graffiti,” Bonadio said. In his study, there is a common comparison among artists. Many artists used allegories comparing AI with food. They described AI-generated graffiti as a ‘cookie-cutter’ that produces similar results or likened it to buying a burger from a ‘fast food’ chain versus the effort of making a meal yourself.

Censorship was also significant issue. “I’m more concerned about being censored than people stealing my work,” one artist commented. Bonadio explained how AI censors swear words, raising concerns about the impact on the thought-provoking and political nature of graffiti.

Bonadio concluded by emphasizing the potential of AI to train students and the desire of graffiti artists to be asked for permission before their art is used by AI:

This question of consent is crucial when working with street art and graffiti.

Giulia Priora, Assistant Professor at NOVA School of Law, Lisbon and Director of NOVA IPSI Knowledge Centre, talked about AI and the public domain. She teaches intellectual property law, legal research methodologies, and EU digital policies. Her research focuses primarily on copyright law, IP theory and functions, and distributive justice principles for inclusive and sustainable innovation.

During her presentation she gave definitions of copyright while showing an image of the roots of a tree, because “copyright has been expanded,” she said. Another definition was for the public domain as “the public’s ability to use words on equal terms without seeking permission.”

She mentioned that there is still a need for better understanding of AI in the field:

"There is a lot of confusion because many lawyers still don’t know what AI is,” she said. “AI was referred to as software but now we lost the definition, and we refer to it as a machine-based system. AI is mostly software,” she continued saying.

Priora thinks that AI can change the law and “not only be a momentum but the moment", helping to use the public domain more seriously.” She encouraged the use of Public Domain works and AI training in education outside the need for copyright authorisation.


Participants in the workshop included:

Ryan Abbott (University of Surrey, UK)

Pablo Fernandez Carballo-Calero (Universidad de Vigo, Spain)

Ioanna Lapatoura (UCL, UK)

Francesca Mazzi (Brunel University, UK)

Andres Guadamuz (University of Sussex, UK)

Caterina Sganga (Scuola Superiore Sant’Anna, Italy)

Noam Shemtov (Queen Mary University, UK)

Philipp Homar (WU Vienna, Austria)

Giancarlo Frosio (Queen’s University Belfast, Ireland)

Peter Mezei (University of Szeged, Hungary)

Joao Pedro Quintais (University of Amsterdam, IViR, the Netherlands)

Antoni Rubi Puíg (Universitat Pompeu Fabra, Spain)

Giulia Priora (NOVA University, Portugal)

Nicola Lucchi (Universitat Pompeu Fabra, Spain)

Zoi Krokida (Brunel University, UK)

Xiao Baiyang (University of Szeged, Hungary)

Magali Contardi (Dirpolis Institute of the Sant’Anna School of Advanced Studies, Italy)

Hasan Kadir Yılmaztekin (Justice Academy of Turkey)

Salvatore Fasciana (University of Sussex, UK)

On Thursday 1 August 2024, City, University of London and St George’s, University of London merged to become City St George’s, University of London. Read more about the merger agreement.

Hashtags