00:00:04:11 - 00:00:32:09 Speaker 1 Hello and welcome to the School of Law podcast from the University of Aberdeen. My name is Neil Whiteman and in this episode we are looking at deepfakes, law, regulation and the growing challenges created by synthetic media. Deepfakes are no longer something that only belong in science fiction or specialist technology circles today. Audio, images and video can be created or manipulated in ways that look and sound increasingly real. 00:00:32:09 - 00:00:59:07 Speaker 1 And I'm sure by now, anyone listening will have come into contact with this kind of media on a fairly regular basis. That raises difficult questions for lawyers, regulators, platforms, creators, performers, consumers, and for anyone trying to work out what can still be trusted online. For this episode, I was joined at the race library by Doctor Rosanna Ducato, one of the senior lecturers here at the University of Aberdeen. 00:00:59:09 - 00:01:28:14 Speaker 1 We also hear from Professor Duncan Mendes, one of the professors at Bournemouth University, who answered the same set of questions separately and shared her responses with us. Rosanna and DeNucci have worked together on deepfakes related research, so it felt right that we invite them both on to this episode. As a quick note. This episode was recorded in January 2026, and given how quickly this area is developing, parts of the discussion may have moved on by the time you are listening. 00:01:28:15 - 00:01:39:21 Speaker 1 To begin with, I asked Rosanna to explain in simple terms what a deep fake actually is and why is it become such an important issue for law and regulation. 00:01:39:23 - 00:02:08:21 Speaker 2 So the term deep fake comes from deep learning and fake, and it means exactly those two words. So it's a digital content. It can be audio, video, images. That means that has been fabricated using deep learning which is subset of artificial intelligence. Now the issue is that that content appears to or I mean looks or sounds to be real and authentic while it's not the case. 00:02:08:21 - 00:02:36:17 Speaker 2 So that thing never happened in in reality. Now, 1st May think that this is nothing new under the sun. We have been knowing impersonators Photoshop, so there is nothing new there. This is, in a way, true, and it's not necessarily a problem, because from a legal perspective, it means that we don't have to reinvent the wheel. What is new and whether is the scale and the impact of the phenomenon. 00:02:36:19 - 00:03:25:17 Speaker 2 The fakes have become extremely realistic. They can be fabricated very quickly, easily by anyone without technical skills. So potentially one with a simple internet connection can create a deepfake and share it worldwide so their impact can be much more serious and much more widespread. Now you ask me about the term deepfake, which comes with a negative connotation. And in a way, this is historical, because if we think about the origin of the term, because it was created in 2017 by a Reddit user who actually used the term deepfake as Nima to post pornographic videos featuring celebrities on on the platform. 00:03:25:17 - 00:03:53:18 Speaker 2 So in a way, it's almost natural that when we think about deepfakes, our mind immediately go to harmful or abusive uses. But that's only part of the picture. The same technology can be used for legitimate and valuable purposes. Think about satire. Parody, other artistic expressions, any form of freedom of expression, really, or tools that can be used for people with disabilities. 00:03:53:18 - 00:04:04:00 Speaker 2 So some of these deepfakes or synthetic media or digital replicas, as we want to call them, can be totally legitimate. 00:04:04:02 - 00:04:13:12 Speaker 1 So explain to us then, why deepfakes can be harmful, especially when it comes to influencing people's opinions or spreading misinformation. 00:04:13:13 - 00:04:46:09 Speaker 2 If fakes can be harmful because by definition they represent something that is fake. So something that never happened, but they are made to appear real, authentic. So that makes them the perfect vehicle for spreading misinformation, influencing elections, manipulating consumers. Imagine you're watching a video of a famous financial expert, and they are telling you that they've just both a certain stock, and it's a reliable and safe investment. 00:04:46:09 - 00:05:23:15 Speaker 2 Now, some consumers could be influenced by that. So they might be they might take that advice. And unfortunately these kind of frauds happened in practice already. Another issue is about the timing of the of the deepfake, the timing of the release of the deepfake. The classic example is deepfake, in which a politician is saying something deeply inappropriate when the polls are still open, even if it's possible to respond quickly to to that thing. 00:05:23:17 - 00:05:33:12 Speaker 2 Releasing a statement, taking down the deepfake, the damage might have been already done because some voters might have been influenced. 00:05:33:14 - 00:05:56:05 Speaker 1 Rosanna. Thank you. In the meantime, we'll hear more from you a little bit later. But right now, I wanted to find out more about how the law is responding. So let's bring in Professor Danish Mendes from Bournemouth University, who has worked with Rosanna on deepfakes related research now. I asked her whether current laws are doing enough or whether they are struggling to keep up with the technology. 00:05:56:09 - 00:06:31:06 Speaker 3 I guess, you know, there's a bit of both because we have some laws that can deal with defects, but in essence, it is a jigsaw puzzle of a variety of laws. And these laws range from intellectual property laws to data protection laws to deformation law, to criminal law, and many others as well. So in other words, there's no single law that deals with defects or one single statute that deals with defects. 00:06:31:09 - 00:07:04:07 Speaker 3 Having said that, the Online Safety Act, which came into force in 2023, which regulates platforms and moderates content, as well as the more recent Data Use and Access Act, which came into force in 2025, is still coming into force, actually, because it has been coming into force through a staggered approach. These two acts deal with deepfakes, particularly in the context of non-consensual sexually explicit images. 00:07:04:08 - 00:07:40:21 Speaker 3 So, for example, earlier this year, croc, which is owned by X, came under scrutiny after it was used to spread non-consensual sexual images of users on the platform. And this involved both adults as well as children as well. And what was happening was that the AI tool was being used to undress individuals, like adults and children, from using images or videos that had been shared on the platform. 00:07:40:22 - 00:08:08:24 Speaker 3 So a very, very distressing situation. But the issue at the time, when it happened in January 2026, was that the relevant section in the Data Use and Access Act, which could have dealt with this particular issue, was not yet in force. Since then, actually it has come into force. So the good news is that there are some laws that can deal with deep fakes. 00:08:09:01 - 00:08:50:14 Speaker 3 Like the example I've just given. However, due to the jigsaw nature of the various regulations, there are also limitations as well. Another. Another limitation in the UK is that of personality rights, which an A personality right basically exists to safeguard creators and artists from misappropriation and false endorsement, protecting their voice, image, name and likeness. Now there are other jurisdictions which has this right, such as USA, Japan, India and more recently I would say Denmark. 00:08:50:16 - 00:09:21:12 Speaker 3 But at the same time, it is not a panacea for solving issues raised by deepfakes. If, for example, UK were to introduce this right, it will need to be balanced with freedom of expression. And a further concern is that individuals could be pressured into signing away any personality right to a dominant player, due to the inequality of bargaining power in the creative industries. 00:09:21:14 - 00:09:33:23 Speaker 3 So whilst we may look to reform existing laws or in fact introduce new laws, it it should certainly be done with caution. 00:09:34:01 - 00:09:43:08 Speaker 1 Can you talk us through a real example or case involving deepfakes that helps show why they are such a legal problem? 00:09:43:09 - 00:10:32:15 Speaker 3 I would draw from a project that I recently conducted with Doctor Rosina Ducato from the University of Aberdeen, and also Professor Tatsuya Uno from Waseda University, Tokyo, Japan. And this project was funded by the Anglo-Japanese Foundation, and it explored the challenges and opportunities presented by deepfakes, was exploring the policy and legislative responses in the UK, EU, Japan and South Korea, and in carrying out this project, we hosted two roundtables, one in Japan, in Tokyo at Waseda University and the second one at Bournemouth University in UK. 00:10:32:17 - 00:11:09:10 Speaker 3 Stakeholders from the technology, policy, academic sectors, as well as those from the creative sector attended our roundtables and we heard many real life examples or scenarios as a result. So I'll give a couple here. For example, Liam barred from equity UK, which is a trade union that represents performers, highlighted the growing concerns faced by performers who fear that their work may be exploited to train current or future training algorithms. 00:11:09:12 - 00:11:39:13 Speaker 3 For example, he gave this example during the roundtable where he talked about how Stephen Fry's recordings of seven volumes of Harry Potter novels recorded back in 2023, was utilized to train and produce a deep big audio clone of the British TV presenter for use in the Hay Festival 2025. But this was something that was unknown to fry and that is pretty scary. 00:11:39:15 - 00:12:41:21 Speaker 3 An example I can draw from Japan is from the, an example that was given by Megumi Moriarty, who is the president of the Association of Arts Workers in Japan, and she stressed how digitalization and defects can impact a working environment which is already strained due to freelance contracts and minimum wage not applied. So she presented a recent survey conducted in Japan which demonstrated that 93.8% of performers feared rights violations by artificial intelligence and 91.9% worried about the unauthorized use of their face, voice or works, and almost 3000 or just over 2000 respondents reported infringements of the intellectual property and moral rights. 00:12:41:23 - 00:13:17:20 Speaker 3 So she further noted that another survey showed that only 2.6% of music and video performers have written contracts which have been enforced, which means, in other words, it leaves many rights unprotected. So these are some examples I can give, which shows that how deep fakes create a legal problem for those in the creative industries, but also other industries as well. 00:13:18:01 - 00:13:27:01 Speaker 1 What makes it difficult for governments to regulate deepfakes without stopping useful or creative uses of the technology? 00:13:27:04 - 00:13:56:06 Speaker 3 This goes beyond defects. And in fact, this is something. This is a question that has been asked all through human history. As long as there has been new technologies. There's always the and this is always a challenge with new technologies, that is knowing when to regulate so as not to stifle new technologies and the creative use of these new technologies. 00:13:56:08 - 00:14:42:24 Speaker 3 It is really important that we strike this balance that is, you know, not regulating too early, because if we do that, then clearly we will stifle the new technology, but then also not waiting too long, because if we do that, then then the new technology is going to become so pervasive and become such an issue for for individuals and all those who are reliant on the technology, whether you are a creator or a user, that it becomes almost difficult then to regulate and to kind of, you know, stop the harm or to to limit the harm from occurring. 00:14:43:01 - 00:15:15:18 Speaker 3 And we have seen this in many different industries and over many, over many years. We can think of that from the perspective of the music industry. You know, back in the back in the day, in the late 1990s, you know, we had Napster, which was a platform that people could go to to access music videos, for example, that were not which have been which were there without the consent of those artists. 00:15:15:21 - 00:15:41:05 Speaker 3 So that was the real kind of, you know, where piracy, music piracy really started. Of course, there was music piracy before as well, but with the introduction of the internet and the and something like Napster, that became a real issue. And of course, for the longest time, we didn't take too much notice of it. It was, of course, very early days of the internet as well. 00:15:41:06 - 00:16:16:19 Speaker 3 Probably we didn't understand the impact that it would have. And then by the time we tried to regulate the law and we tried to regulate copyright law, it became quite difficult. And ultimately it was technological solutions together with the law that like, for example, through tools such as iTunes or Spotify or in the case of films like Netflix, that has allowed the user to access that material at a reasonable price, and not then to enter into the world of piracy or in fact, in the case of trademarks, counterfeiting. 00:16:16:19 - 00:16:55:08 Speaker 3 So this is also the same challenge for defects as well. AI technologies are used intensively, intensively for for example, in the sciences, in the, in the medical field, for good things, for positive things, for treating patients. And we cannot stifle the technology because, for example, you know, it also has some very bad negative consequences. But the thing with defects at the moment is that the regulation has been left in the hands of online platforms, which use automated means to take down deepfake content from the internet. 00:16:55:08 - 00:17:21:23 Speaker 3 But the problem is that algorithms do not know or understand the context in which those images are taken down, and also in some cases, maybe it can lead to it can lead to censorship, depending on what type of deepfake it is. Is it is it a kind of a parody that is not harmful? Or is it is it maybe a type of content that's been used for education? 00:17:22:00 - 00:17:55:14 Speaker 3 For illustration, in the case of teaching. So it is a challenge. And like I said, we have seen this through many years, all from the moment that humans started inventing new technologies. In the case of copyright, we can go back to the 1400s when the printing press was created and invented. And that was the first time that we that someone was able to recreate the same work without having to handwrite the new text. 00:17:55:18 - 00:18:20:17 Speaker 3 And then from that moment, from that moment of the printing press to the fax machine to the for copy machine, to the internet and now to deepfakes, you know, this question of when is it that we regulate and when is it we do that without stifling the new technology? This question has been asked for many years, and the governments are now having to deal with the same question. 00:18:20:17 - 00:18:23:02 Speaker 3 But in this context of deepfakes. 00:18:23:02 - 00:18:49:10 Speaker 1 Professor Dinesh Mendes from Bournemouth University, thank you very much for taking part in this episode of the School of Law podcast. We are, of course, talking deepfakes. With me now is Doctor Rosanna Decatur. We heard from you on. I'd like to ask you another question. Who should be responsible when a deepfake causes harm? The person who made it, the platform that shared it, or someone else? 00:18:49:12 - 00:19:23:24 Speaker 2 Well, if the deep fake is harmful. So we're not dealing with satire, parody, or other legitimate or authorized used in principle, anyone, everyone who played a role should be liable. Now, the current law already establishes some clear responsibility for people who created or shared an illegal or harmful deepfakes. For example, we were mentioning the non-consensual intimate deepfakes, but it could be liable also, the platform who is ultimately monetizing that content? 00:19:24:01 - 00:19:55:19 Speaker 2 And it recently made the ad lines that the Ofcom has opened an investigation into grog for the violation of the Online Safety Act. And because it has been reported that there chatbot has been used to create undressed versions of pictures and videos, unfortunately usually of women and children. So I think that this is investigation is already an encouraging sign showing that the law is there and we are testing it right now. 00:19:55:21 - 00:20:06:02 Speaker 1 I wonder then, could you give us an idea? Looking ahead, what should students expect the law on deepfakes to focus on in the future? 00:20:06:04 - 00:20:35:22 Speaker 2 Lawyers are essentially problem solvers. The fakes, like any other emerging technology, raises complex challenges. That requires right now, multidisciplinary or interdisciplinary approach. After all, if we do not, then how the technology works, how can we regulate? Or if you're a lawyer, if you are a professional, how you can provide advice if you don't understand the basics, if you're on a stand, one might be the consequences the same if you are a judge. 00:20:35:23 - 00:21:05:11 Speaker 2 How you can decide a case which deals with with deepfakes. So I think that interdisciplinary will be more and more critical for for law students and to legal education will will go in this in this direction. Now I really like to use this image to explain this kind of of trend or how legal education could evolve. In a way, this image has been already used in the literature. 00:21:05:11 - 00:21:36:18 Speaker 2 So it's not I'm not making anything up. It's the image of the T-shaped lawyer. So like the T, the vertical line, you are solidly grounded in your own field of expertise. And the Tea Party, the your line is like your arm open to others reaching other, other expertise. Now, lawyers are not supposed to become computer scientists, sociologists, experts in all possible fields. 00:21:36:20 - 00:22:22:13 Speaker 2 Perhaps someone will. But the important part is that they should be able to understand and speak the language of others. And this is, in a way, what we are trying to do. Also, at our own university. My colleagues and I have been recently awarded an Erasmus Plus Year monitoring for Digital Transition, and as part of this chair, we will offer a series of interdisciplinary courses where experts from computer science, psychology, philosophy will share their own knowledge with our law students to understand certain topics that are raising some interdisciplinary issues better. 00:22:22:15 - 00:22:52:09 Speaker 2 And then in particularly proud, because we will have a so-called legal design clinic where students in law and computing science will work in team to solve a practical problem and create solution for how can we create an AI tool that is compliant with the regulation? Or how can we provide, in simple and understandable terms, privacy policy from an IOC system. 00:22:52:09 - 00:23:15:13 Speaker 2 So these kind of challenges, they will have to learn the law. They will have to understand the technology. They will have to put that knowledge in practice in mediately and create something new and original. So in a way, we are trying to simulate what they might see in their future career, but in the safe environment of the classroom. 00:23:15:15 - 00:23:32:02 Speaker 1 You know, I've really been fascinated with this chat, and I thank you for your time. Before we call a close to it, though, I would just like to ask if someone listening to this podcast is fascinated by deepfakes and the subject that we've been talking about today, what does the University of Aberdeen offer to its students? 00:23:32:04 - 00:24:06:15 Speaker 2 Well, we have a course specifically dedicated to low energy, so it's a modular course where we explore different aspects of the law that have been innovated or that have been challenged by the rays of artificial intelligence. So, for instance, we are considering the face, but also the intellectual property aspects of artificial intelligence, for example, to what extent is legal to train your algorithm on copyrighted protected materials? 00:24:06:16 - 00:24:33:07 Speaker 2 Or what about the output of that process? Is it protected by copyright? Is it not? Who is the author? We are even challenging the notion of authorship right now. So there are a lot of interesting issues there. Or also in terms of liability, what happens if something goes wrong? Who is responsible? The developer? The platform, the user, the end user. 00:24:33:09 - 00:24:38:07 Speaker 2 So our students are now confronted with all of these interesting questions. 00:24:38:09 - 00:25:13:07 Speaker 1 Doctor Rosanna Decatur, thank you so much for joining us in this nice library for this episode of the School of Law podcast. Thank you, thank you. So there you have it. Deepfakes, the latest episode of our School of Law podcast from the University of Aberdeen. My name is Neil Whiteman. Before I sign off, I just wanted to say a big thank you to our guests, of course, Doctor Rosanna de Cato and Professor Duncan Mendes, but also to you for listening and anyone who subscribed to the podcast or given us feedback or indeed share it. 00:25:13:07 - 00:25:19:19 Speaker 1 It just helps us get seen and heard by more people who might enjoy it. So thank you. A big shout out to you, 00:25:19:22 - 00:25:24:09 Speaker 1 right? That's it from me until the next episode. I only have to say one thing goodbye.