 Hi, thank you so much. I'll just share my screen. And yes, say again that I'm extremely pleased to be here. So thank you so much for the opportunity to speak at this conference. My name is Brigitte Liesner, and I'm the director of policy at Creative Commons. So my background is in copyright law. And so it's from that perspective that I'll be speaking to you today about open content, but more specifically creative commons license content and artificial intelligence. So there are actually many questions that we need to find answers to and many problems that we need to find solutions to, but I'll focus on one today. And that can or should openly license content. And by that I mean photographs or artworks, text or music, especially creative commons license content, should it be used as an input to train AI. And to answer this question, we have to take a look at the CC licenses and tools themselves. But even before that, let me share a real life example. Maybe some of you have heard of it. In 2019, we at Creative Commons learned that researchers at companies like IBM were training their facial recognition AI programs by feeding their algorithms with CC license photos that were publicly available on Flickr, for example. So I think there was about one million, sorry, let me go back. There was about one million photos taken from Flickr website used to train facial recognition program. Those photos were openly accessible through Creative Commons licenses. And IBM had not told or even asked permission from the people that were either photographed or from the photographers. However, some Flickr users were dismayed to learn that IBM had used their CC license photos to train the AI, all the more so as it was done for commercial advantage. They had questions about the ethics and the privacy implications of such a data set being used for algorithmic training. So the question was, well, do the terms of the CC licenses allow the open content to be used as input to train AI without any further permission from the licensor? And to get a sense of the various views on this question, we launched a Twitter poll back in January and nearly half of the respondents says, it depends. And we agree. And here's why it depends. I'm hearing a bit of background noise and maybe it's just me. Yes, better. So on the one hand, Creative Commons is dedicated to facilitating greater access and openness for the common good. So we believe that the use of openly accessible content can lead to greater innovation, collaboration and creativity. We also believe that the limitations within copyright law, which generally privilege their reuse of the facts and the ideas that are embodied in creative works contribute to a rich and generative public domain. So we support, in principle, broad access and use of copyright works, including openly licensed content to train artificial intelligence systems, because such access can, for instance, help reduce bias. It can enhance inclusivity and inclusion. It can promote important activities like education and research and it can also foster beneficial innovation in the development of AI. But on the other hand, with new patterns of online sharing billions upon billions of copyrighted works that have become the basis of business models, including those that are built about processing all that content turned data. Copyright licensing becomes inextricably linked to privacy and ethical concerns and something that the licenses were not originally designed to address. So as people share vacation photos online, for example, on Flickr, they're contributing to a massive trove of data that's processed by potentially, you know, a lot of machine learning or algorithms. And I should note that this is likely to be the case whether or not these photos are shared using CC licensing or not, because in many jurisdictions, at least in the United States, copyright has very limited purview of machine learning. Also, to the extent that copyright can be leveraged to restrict invasive machine learning over content, Creative Commons is reluctant to kind of encourage weaponizing copyright law for purposes that are far outside the original goals of copyright. So, I think it's important to recall at this point that Creative Commons licenses are copyright licenses. They're designed to give creators of all types of work a simple way to grant public permissions to use their works. These tools were, or licenses and tools, they were a response to overly restrictive copyright laws in the context of a digital ecosystem where sharing and remix were easy and ubiquitous but where copyright was imposing unnecessary barriers to the exchange of knowledge and culture. So what does this mean in the context of AI training? Well, quick refresher about the CC licenses, three main points. The first one is that our licenses, they don't restrict the reuse to any particular types of reuse or technologies. So long as the terms of the license are respected. So that's the attribution requirement, so CC by, the share like or SA, no derivatives or ND and non-commercial NC. So from a strictly copyright perspective, there's no special or explicit permission that's required from the licensor to use the CC license content to train the applications to the extent that copyright permission is required at all. And that brings me to the second point, which is that the licenses do not override limitations and exceptions, like fair use. So if a use is not one that requires permission under copyright or sui generis database rights, for example, if it's, if it's text and data mining that's allowed under an exception to copyright, one may conduct the training activity without regard to the CC license. And the third point is it's also report important to recall that our licenses operate within the copyright system. So privacy, personality, publicity and other types of rights or even ethical considerations are not covered by the licenses. They're not universal policy tools there, they're limited to the confines of copyright law. So copyright is the primary obstacle to reuse that the licenses solve, but there are many other issues that are related to the reuse of content that our licenses do not address, and that re users have to be aware of and like I said this includes privacy rules governing ethical rights, the search, the collection, the use of data data protection, and these have to be addressed and respected separate and apart from the copyright issues that the licenses cover. So we do our best to ensure that those releasing their creations under CC licenses understand the scope of the copyright right that are managed under the licenses. License may offer users permission to reuse a photo. It would not offer the permission to make the use of the personal likeness of the people in the photos, which might be governed by image or personality rights. So whether one has to comply with the copyright regime, and hence with the CC license terms will depend on whether the type of AI training activity is an exercise of a right reserve to the rights holder. This is a very useful workflow if you if you really are facing this question, I encourage you to take a screenshot right now I won't be going through every single tree branch but I think it's a good representation of the many situation what might one might encounter. And one of the, the main difficulty is that there remain significant legal uncertainty about whether copyright applies to AI training, which means it's not always clear whether a CC license applies. So, in other words, there is no consensus on whether the use of copyright words as inputs to train an AI system is an exercise of an exclusive right, right it doesn't involve reproduction of the work does it. Does it involve its adaptation, etc, does it trigger some kind of exercise of an exclusive right. I'm happy to go through this in the Q&A if we have time at the end, by the way. I said the situation is likely to vary across jurisdictions because countries are now progressively regulating the copyright AI nexus. Like I mentioned in the US the use of words to train AI is likely considered fair use. In my view, there's article three and four of the DSM directives that I would refer to that's on text and data mining that that foresees a regime that's a bit different, whether the TDM is done for commercial or non commercial purposes. As a matter of copyright laws of the, we believe that the use of works to train AI should be considered not infringing by default, assuming that access to the works was lawful at the point of input. So, to give you an example, if someone conducts text and data mining in the context of research or education and that's allowed under an exception. So, if you do the adage that the right to read is the right to mine then the license would not be triggered. However, we do recognize that today, every act of online sharing potentially implicates privacy or ethical concerns. And some believe that this is an inevitability that foregoing personal privacy is necessary when sharing content online but we still imagine a world where sharing and privacy can coexist and that this coexistence is paramount to sustaining free and democratic societies. And this is why our new strategy. I have a little excerpt up there on the screen. It recognizes that we need an approach that transcends copyright. So the conversation on artificial intelligence within copyright policy must be held also in a coordinated and inclusive manner, and through the lens of ethics, responsibility sustainability, also cultural rights and human rights and personality rights and privacy rights and data protection. As with any fundamental ideal, the openness of data is should not be considered an absolute end in and of itself, but it has to be balanced with equally valid considerations the ones that I just mentioned to ensure that sharing ultimately benefits the public right we want to promote better sharing for a brighter future, and that might be where the future of open is so beyond copyright issues. Artificial intelligence is likely to affect the sharing of creative content and the open community in general. This legal uncertainty that I mentioned caused by the ethical concerns about AI, the lack of transparency of AI algorithms and the patterns of privatization and enclosure of AI outputs. All together, they constitute yet another obstacle to better sharing because for many creators, these concerns are reason not to share. So while we do support broad access to content, we also aim to increase our understanding of the ethical concerns that may constitute barriers to open sharing. So I would invite you to join this conversation. We want to promote promote the use of CC license content to train AI, but we think that we need a community led coordinated and inclusive approach to not only consider the copyright system, but also these other issues of responsibility accountability sustainability, etc. So we're one actor in a vibrant community of open advocates that promote the interests of the millions of people who use CC licenses, and we want to engage in rich conversations on AI multiple facets to promote better sharing in the public interests. And to that end, the Creative Commons copyright platform working groups have been set up. And throughout the year, they will examine the intersection of AI and open content as well as the ethics of open sharing. And through these discussions and and collective action while we look forward to identifying options in licensing and infrastructure policy norm building, and also awareness raising. So I would invite you to join them there are two that you might be interested in AI and open content, or the ethics of open sharing. And they will they will be presenting the outcomes of their work at a public webinar on November 9 so if you're interested, please do mark your calendars and get in touch for any further information. So with this, I thank you for your attention and I look forward to any questions that you may have. Thank you. Thank you so much for this presentation, Bridget. We do have questions and I think we have at least time for one. So Jeremy asks, does it seem from your user community that there's interest in deriving a new CC license for no machine learning or no artificial intelligence. I think this is one of the options that being explored is to see whether there is a need for to develop a new type of license that would regulate, or that would give permissions based not on copyright but on ethical considerations. That's one option that's being looked at at the moment. Another one would be that there could be an ethical layer that's built into the existing licenses so not a separate license but that each license would include some some kind of layer in addition to the existing license that would take into account privacy or ethical concerns. But these are these are all open questions that we are beginning to explore. And I think that if you if you'd like to learn more, the working groups are probably the best way to do to do so. Alright, thank you. And another question. What do you think about the development of structured data in wiki data used as basis for commercial applications that is machine learning plus crowdsourcing. Yeah, I don't have any specific knowledge about that. Those data sets. But I think that yeah, on the principle creative comments would support access to open data to the largest extent possible because we do think that there are many benefits to using as diverse and and very data. But at the same time I want to point to these limits that I that I mentioned where I think that openness maybe stops to be the driving focus and that there are other perhaps equally important considerations that need to be taken into account. Right. In the last question because we have time for it. And Lee asks, how do these licenses apply in the context of other data that is non creative like geolocation and is thinking of open street maps here. Yeah, no that's a very important question and the creative commons licenses only apply to copyright content so if it's something that's not copyrightable so if it's not original in the copyright sense and does not constitute a copyright work. One cannot attach a creative commons license to it so we wouldn't advise applying any creative commons license on on material that's not copyright protected.