Navigating the AI Legal Frontier

[Music] uh Rachel the answer to your question as to who owns the IP uh for your resolve pop song is probably nobody um there's no bill for that uham alaykum good morning good afternoon your excellencies your honors justices ladies and gentlemen all curious Souls Linda highlighted the fact that we're here focused on assets investment and asset management we've seen eye popping Investments and heard hyperbolic claims about the potentials and dangers of artificial intelligence I'm going to talk about a slice of that world the role of Law and courts in the development and promulgation of this promising technology and I hope you enjoy the topic of the talk I'm going to focus on copyright regulation and AI specifically Global being the theme of Abu Dhabi Financial week and maybe Abu Dhabi in general the local and the global and if you don't enjoy the talk it's 1:00 a.m. back home in New York City the Advent of generative artificial intelligence has ignited a global transformation one that promises both unprecedented opportunities and profound challenges Nations such as the United States China members of the EU and here in the UAE are each racing to advance and regulate this technology shaping the future of Innovation and governance yet at the heart of this debate lies an intricate and Urgent question how do we reconcile existing rights such as copyright law with the realities of AI development especially when it comes to training these powerful Technologies I assume most people in the room are familiar with how AI models are trained if not here's a quick overview AI models based on their architecture generally speaking become more proficient based on the volume of data on which they're trained similar to how children learn based on the more experiences they're exposed to AI models are able to predict or infer as we use in the parland what we ask of them better when they're exposed to more data but where does this data come from answer is books the internet basically all of the written artifacts of humanity so since we're talking about training on a huge Corpus of data as much as data in these very large language models can acquire we have to think about the issue and use of copyrighted materials for AI training this practice which lies at the foundation of generative AI has sparked a spectrum of responses worldwide and frankly has taken courts and legal uh practitioners relatively by surprise for doing a survey of the world it falls into three categories there are jurisdictions with fair use provisions like the United stes states where I practice where courts often allow copying for purposes that don't communicate the original expression to the new audience there are jurisdictions with Express exceptions for text and data mining like Japan and the EU which expressly carve out rights for AI training under very specific circumstances then there are countries with aspirations for AI dominance but have different and older copyright Frameworks such as China including uh the the UAE as well which are yet to fully address those challenges in their laws I want to point out that in the UAE article 22 of the law on copyright and neighboring rights does allow for the reproduction of a single copy of a work for purely personal use and for nonprofit and nonprofessional purposes despite these differences in laws we're seeing an emergence of what can be called a global equilibrium Amore is around recognizing that AI training and similar computational analysis are socially valuable and don't inherently harm the rights of copyright holders taking a step back and at a higher level at its core copyright was designed in response to the printing press emphasizing the reproduction of works as a central concern but as generative AI flips this Paradigm AI systems are relying on non-expressive uses of copyrighted Works which are basically acts of copying that no human will ever see or read in order to teach a machine patterns abstractions and relationships take the training of large language models like gpts these systems ingest huge amounts of data transforming the text into new uh numerical representations called tokens and even though this involves copying it serves an entirely different purpose than what we're used to it's taking uncopyrightable facts and insights rather than just merely reproducing the works on which they trained they isn't Napster so this raises a crucial question should copyright law permit these uses particularly when they unlock incredible potential for social progress going back again to the goal of copyright recall that it includes the idea that the author has exclusive rights to reproduce the work but those rights are not absolute it can't be if it was life would be exceedingly boring and we'd have little to no scientific or artistic progress imagine if there was only one impressionist painter or there was no satire at all be a very dull World indeed so at the core of copyright is this idea of protection but also progress the idea that the author controls communication of their original expression but the underlying facts ideas abstractions artistic methods and yes even style can be freely shared and copied so with that that in mind I want to highlight three factors driving this convergence and and This Global alignment how countries are thinking about AI the first this core idea and copyright of the idea expression dichotomy this idea that while expressions are protected the ideas are not second is global AI competition we've seen the jockeying between nations for AI Supremacy and as Nations VI for leadership in AI they look at the flexibility built into US copyright law and in particular the fair use Doctrine which I'm litigating in several cases this has placed a pressure on others to adapt countries like China the UAE the UK and others are looking to see what's happening in the US this is then driving this race to the Middle where other countries are looking for a core of moderation so even though we have this consensus that's forming around the middle the emerging consensus is fragile people people are looking to litigation in the United States debates over licensing deals and legislation in the EU such as the EU AI act to possibly disrupt this balance there are questions over whether even the outputs of AI such as uh the point Rachel mentioned right the the outputs whether they resemble copyrighted Works would that be an infringement does it go to the public domain what does it mean I think these are all being the questions being asked in courts and are yet to be fully litigated looking ahead there are six lessons that stand out that I want to leave you with around this Global survey first is that balance is key effective and strong copyright regimes have to navigate the fine line between promoting Innovation and protecting rights second non-expressive uses also deserve recognition in the law a certain type of copying that enables AI training without reproducing the underlying creative and expr Ive elements of the authors should probably be permissible the third flexibility Fosters leadership jurisdictions with adaptable laws like the United States are better positioned to lead in Ai and we've seen that reflected in investment fourth Global collaboration is essential AI is inherently International and harmonizing copyright laws will reduce friction and enhance Innovation five watch out for overregulation excessive restrict on AI training could stifle progress and disadvantage nations in the global race we've seen that acknowledgement with countries like Japan as well as the UAE which are seeking to be leaders in AI last engagement with stakeholders policy makers must actively involve copyright holders AI developers and the public in shaping the rules of this new era ladies and gentlemen the intersection of copyright and AI is more than a legal challenge it's a defining moment for innovation creativity and Global cooperation and as we navigate these complexities let us remember that the goal of copyright is not to hinder progress but instead to enable it with thoughtful policy and collaboration we can build a framework that supports the creators of today and the pioneers of tomorrow thank you very much [Music]