The Network Law Review is pleased to present a special issue entitled “The Law & Technology & Economics of AI.” This issue brings together multiple disciplines around a central question: What kind of governance does AI demand? A workshop with all the contributors took place on May 22–23, 2025, in Hong Kong, hosted by Adrian Kuenzler (HKU Law School), Thibault Schrepel (Vrije Universiteit Amsterdam), and Volker Stocker (Weizenbaum Institute). They also serve as the editors.
**
Abstract
As the levels of speed and control of Generative AI tech increase to a degree that any media can be changed in real-time to the whim of the consumer, content is becoming interactive – each image its own canvas, each song its own instrument. Counter-intentionally, copyright appears to drive this new form of production towards large prosumer social media platforms, given they may be the only spaces that can legally support the mass of copyrights necessary for widespread immediate media regeneration. If works can easily be regenerated into similar shareable versions, strict copyright enforcement may incentivise creation of “rights-washed” public domain versions which are more amenable to new forms of interactivity, further pressuring rightsholders to allow remixing of their works to maintain attribution. Thus, copyright’s friction with new media may embolden a market dynamic that in certain instances undermines its own value for rightsholders, reliant on control mechanisms in an era increasingly defined by uncontrolled spread and seemingly infinite remix.
*
1. Introduction
Copyright is traditionally justified on the basis of incentivising innovative production.[1] Yet, this rationale for copyright has long come under fire, as sceptics have pointed out instance after instance where copyright appears decidedly removed from stimulating production.[2]
With the advent of Generative AI (GenAI) works, we see this rationale once more stretched, as GenAI allows massive amounts of production with minimal investment and indeed, entirely innovative new forms of production, yet the majority of copyright frameworks worldwide have roundly denied copyright to AI-generated works. Although this has been under the auspices that copyright is here to incentivise human production, this rationale has only ever been brandished against human authors.
This misguided debate over whether human authors have authored their works has distracted from copyright’s emergent relationship with new forms of GenAI-enabled media in contemporary creative markets and the likely impacts of denying them copyright.[3] Historic assumptions around foundationally different production and consumption conditions have failed to shed meaningful light on copyright’s more complicated relationship with contemporary AI-production. Newly interactive media have their own intricate logics as do the platforms where they are likely to proliferate that are broadly in tension with the logics of copyright, built on controlling use of your works for economic benefit.
Already, individuals can feed songs, videos or photos into GenAI tools and have them near-instantly remixed and repurposed within seconds, not minutes. Yet copyright scholarship has roundly neglected the implications when the levels of speed and control of current-state technology increase to a degree that any media can be changed in real-time to the whim of the consumer. This would render media instantly interactive, meaning each individual would become an artist inasmuch as they have interest in changing the content they consume – each image its own canvas, each song its own instrument. We are flying into the age of “can I hear this Bob Dylan song as a funk song but with a female gospel singer, please?”[4]
In the meantime, platforms are voraciously fed content. 16,000 videos are uploaded to TikTok every minute.[5] Some of these are expressly artistic while others are social. Which are which become harder to delineate as even the most basic video for your friends is easily soundtracked, filtered, animated, transformed, elevated. The professional and the amateur sit on a spectrum, not a binary – each schlub holding the lottery potential to go massively viral for something they created, catapulting them to potential monetisation opportunities.
Thus, we are entering an era where the individual can direct any piece of content to transform itself into something else. This may be a watershed moment for creativity – akin to the advent of recorded music – sure to foundationally change our entire production / consumption (prosumption) dynamics for the 21st century.
2. The Legality of Interactivity: Why Social Media Platforms May Win Control
Yet, in order to transform media fluidly, we must have the rights to do so. It is exactly this sort of infinite fluid transformation that copyright stands in the way of. Although copyright frameworks allow us to privately remix protected works if we own the works, increasingly people do not own much of the content they consume – rather engaging with it digitally through platforms.[6] This means that the complex web of rights and private licensing that governs our creative consumption will dominate what we are allowed to do from the first moment of production in this new era of instantly interactive media.
Here, we see an immediate tension between copyright’s ostensible intention and its relationship with new media transformation – namely, it primarily prohibits this new form of innovative production unless individuals are able to acquire broad licensing rights for each work they wish to transform that few will be in a position to. (Naturally, copyright can be defended upon the notion that individuals would not create works if they knew that others could immediately repurpose them without their original creators receiving any revenues. Yet, given contemporary creative platform dynamics, this is not necessarily true for much of the content being created today. We will come to this later.)
Inherently then, for interactive media to function en masse, it must take place within walled gardens where rights to remix (a lot of) the media have been cleared. The most obvious groundskeepers, then, would be social media prosumer platforms such as TikTok, which can easily clear transformation rights over any content native to those platforms in their terms of use. Social media platforms’ architectures are inherently memetic, thriving on the circulation of endlessly repurposed content, and in turn, are naturally designed for mass regeneration of differently versioned media.
These prosumer platforms will have a leg up on consumer platforms that are reliant on third-party licensed media, such as Spotify or Netflix. Although major music streaming platforms broadly aim to be one-stop-shops for as much music as possible, they still do not allow in-app uploading. Rather, all of the hosted music is licensed to the platforms by distributors. . While there are readily accessible online distributors for artists that automate this process relatively painlessly, it remains third-party enabled– a distributor licensing music to Spotify, rather than an in-app “Upload” button.[7] The licenses to the platforms do not generally include any rights for users to remix the works.
Social media platforms also license copyright-protected material onto their platforms, such as major-label music to soundtrack users’ videos. Yet, these licenses also do not generally allow users to remix the content, meaning users could only legally use these new tools on platform-native content. Thus, without renegotiating licenses for distributed content, it will likely be those platforms that already have large user bases, and in turn, large amounts of native content that are best equipped to host instantly interactive media remixing.
If users grow to love interacting with content, this may lead to a surge in excitement for transformable content, which can be melded to the whims of the user, over third-party distributed content, which cannot. In the extreme, such circumstances would mean that copyright would drive interest away from protected content toward openly remixable content. This is therefore a critical open question – to what extent would people be interested in instantly transformable content?
In a sense, this is actually three questions:
1) To what extent are people interested in bespoke content?
2) To what extent do people want to themselves mould this content?
3) and finally: to what extent is that interest for private enjoyment and to what extent is it for creative or social production?
Some users may love sitting in their bedroom listening to a hundred different versions of their favourite song in one hundred different genres with no interest in any of these ever seeing the light of day. We might imagine this as a GenAI-enabled streaming platform feature in the near future. Others may have little interest in doing so whatsoever, yet may see any new hyped content as an opportunity to offer their own spin in pursuit of algorithmic fame and fortune.
Different answers to these questions will have enormous impacts on copyright’s ability to empower rightsholders. Copyright is only as valuable as the commercial options it allows. If a marketplace is dominated by a platform architecture with its own predetermined terms (and rightsholders are not powerful enough to renegotiate), rightsholders can either submit to these terms or try to monetise elsewhere. Thus, copyright is of decreasing value to the extent that no other commercial opportunity exists outside of the platform. It is inherently in the interest of rightsholders, then, that platforms do not hold so much public attention that rightsholders are both beholden to their (likely unfavourable) terms and unable to garner much interest elsewhere. The more that a platform forms and controls the marketplace, the less relevant copyright becomes. Join or perish.
It is therefore inherently troubling for rightsholders if copyright drives this innovative new form of production towards prosumer platforms if that diverts attention away from other more lucrative avenues. Just as rightsholders have traditionally been reluctant to allow users to transform their works if they feel this creates competition against their own work without adequate payment, they may just as readily allow users to remix their works if they feel that prohibiting remix renders their works less competitive on social media prosumer platforms.
3. Rightsholder Pressures: The Need for Visibility and Attribution
This pressure to allow remix rights on social media platforms may be exacerbated to the extent that any commodity and primary value of work has been diminished, such that visibility and fame are more important than directly profiting from any content created. As an example, we might consider the hierarchy of increasing attention to decreasing value seen in music consumption: from high-paying low-visibility physical commodity music, low-paying higher-visibility music streaming services, and then finally, miniscule-payment high-visibility social media platforms. For each tier, musicians accept the lower payment in pursuit of higher visibility (which they hope to monetise elsewhere).
Physical music revenue is in global decline, often now marketed as a form of specialist merch, as music is primarily streamed or enjoyed live.[8] While a study last year found that a massive 70% of European musicians were not satisfied with streaming royalties while Spotify estimated that they only hosted 225,000 professional musicians worldwide with only approx. 67,500 earning over $10,000 a year from Spotify royalties in 2024 (compared with an estimated 11.6 million full-time influencers in the US alone), the vast majority of professional recording musicians continue to license their music out to Spotify.[9] Without any other viable marketplace, many hope that through maintaining platform visibility, they will be able to monetise that heightened brand value elsewhere (such as through live shows).
Social media platforms generally pay even less in music royalties.[10] The memetic focus of these platforms is inherent in their royalty payout scheme, which pays per video that uses the song, rather than per stream of that video. Yet, it remains critical to many artists, for whom presence on these platforms is essential. When Universal briefly pulled its music catalogue from TikTok in May 2024 in pursuit of higher payouts, many of its artists pushed back, no longer able to properly promote their music without adequate social media visibility.[11] “Brat Summer”, the viral phenomenon that generated extraordinary amounts of attention for Charli XCX’s 2024 album “Brat”, could not have saturated social media if users were not allowed to widely soundtrack their videos with Charli XCX’s music.[12]
Critically then, to the extent that content’s core value is not in its own consumption but rather as an ad for value elsewhere via heightened visibility, then its copyright may work against it. Copyright allows rightsholders to control the flow of their content, yet ad value comes in its memetic qualities – the less restricted its flow, the better. In such circumstances, it is attribution that is paramount, in order to monetise the fame elsewhere. This is a critical dynamic to understand when it comes to interactive media. Inasmuch as copyright allows rightsholders control of their work, this right loses value to the extent that content is most valuable in its uncontrollability – inviting users to create their own versions and to freely distribute them widely.
To the extent that this type of interactive prosumption is preferred to passive consumption and to the extent that the content itself has lost primary intrinsic value, the greater the pressure upon rightsholders to allow mass regeneration and remix. This will also be affected by whether the rightsholder actually captures the secondary value of the content (ie if a record label does not profit from an artist’s other revenue streams, then allowing music to disseminate for free to find value elsewhere may not be desirable).
We might contrast this type of content with, say, a high-production-cost Hollywood film, which may only stand to lose value if others can create their own versions. Still, if high-production costs are increasingly difficult to recoup in competition with low-cost instantly remixable works which invite participation and can circulate virally more easily, there is a risk that this could lead to less high-cost production.
Regardless, to the extent that widespread licensing and network effects are necessary for legally interactive media, copyright inherently empowers massive entities to control this new mode of creativity, which may define the century to come. In turn, different rights over different content may create a dichotomised content environment – that which users are allowed to change and that which users are not.
Platforms that have no native content, like Spotify, must then either stand their ground as a space of consumption only, start generating their own native content or renegotiate third party licences. Given reporting that Spotify has been preferencing “fake artist” muzak licensed from stock-music companies in their recommended playlists in order to pay lower royalty rates, directly generating Spotify-created interactive music onto the platform may not be so farfetched.[13] Netflix’s own flirtation with interactive content has for the most part dampened, although its interest in gaming is only intensifying.[14] If interactive gamified content proves just too attention-grabbing, we may find the dividing lines between prosumer and consumer platforms to be less about interactive versus non-interactive, but rather public interactivity (instant content creation) versus private interactivity (interactive videos / gaming).
It thus remains to be seen the extent to which traditionally consumptive platforms and protected works rightsholders will feel pressured to allow environments to interact with their works, thus rendering media accessible to instantly remix upon direction, or the extent to which legacy copyright creates a boundary between active interactive content and passive consumptive content.
If platforms do end up licensing rights for distributed content to more readily flow alongside interactive native content, we may pass a watershed moment in copyright history – as norms are shifted such that users have permission to infinitely remix much of the content they consume, provided they do so within the walls of platforms. If attribution is king in the realm of such infinite regeneration, the denial of copyright to AI-generated works may further disempower rightsholders, as works can be cleaned of their rights through AI regeneration.[15]
4. Rights-Washing & Public Domain Metamorphosis
Currently, public-facing AI tools allow for near-instant remixing of other content (taking seconds, not minutes). At commercial music-generation sites like Udio and Suno, users tick a box claiming that they are not uploading any protected content they don’t have the rights to and can then receive instant remixes of songs – protected or not. Content ID style mechanisms, wherein content uploaded is checked against a database of copyrighted content, are decidedly limited nor are even they necessarily mandated.[16]
US law does not mandate content filtering for hosting services.[17] Although in Europe, Article 17 of the Digital Single Market (DSM) Directive does establish that Online Content Sharing Service Providers (OCSSPs) must make “best efforts to ensure the unavailability” of copyrighted works (widely interpreted as referring to filters and automated content recognition systems), there may something of a loophole at play here.[18] A platform is only an OCSSP if one of their main purposes is to “store and give the public access to a large amount of copyright-protected works”.[19] Since European courts have until now rejected granting copyright to AI-generated material, the material stored could be argued as not being copyright-protected, and in turn, AI-generation platforms would not need to adhere to the filtering requirements.[20] Yet, to the extent that the material generated still incorporated part of the copyright protected works uploaded, those elements would be copyright-protected, in which case Article 17 of the DSM would apply. Although these provisions may then protect against these creation platforms becoming databases of regenerated works, they would not provide adequate protection to stop private regeneration of protected works.
The denial of copyright to GenAI works thus highlights a clear tension – namely, users and platforms alike are incentivised to regenerate works into similar works that are no longer beholden to the restrictions of protected content. Inasmuch as any regenerated content contains elements of the copyrighted content within it (eg the melody of the song was the same but in a new genre), it can still not be legally distributed elsewhere. Yet, to the extent that it no longer resembles the copyrighted content, it would fall into the public domain. Thus, each regenerated work contains a protected core with an unprotected generative runoff. To the extent that that core disappears or substantially changes with each regeneration, the entirety becomes public domain generative runoff in a kind of public domain metamorphosis. Users are now free to do what they wish this newly public knock-off. Indeed, it is not inconceivable that a tool could be trained specifically on copyright law to “rights-wash” works, changing them just enough to no longer be considered derivative works of the original.
While rightsholders might therefore hope that access to these tools can somehow be restricted, new platform dynamics may give pause to cutting off this new form of production entirely at the root. First of all, it is hard to imagine that creation of these works can be entirely controlled. If every Taylor Swift fan is seconds away from creating their own Taylor content, cutting off the flow may prove impossible. This content is going somewhere. Immediate rejection of it from platforms through automated filtering (a la the DSM Directive’s “best efforts to ensure unavailability”) may be unwieldy (as it is not identical to the protected content) while also denying rightsholders the chance to profit off of any that end up being proving popular. Emblematically, YouTube allows rightsholders to monetise protected content that has been uploaded without their permission (itself a private distortion of public copyright logics).[21] Tellingly, rightsholders chose to monetise 90% of all Content ID claims in 2024 rather than remove it.[22] At a certain point, rightsholders may consider whether they prefer granting permission on a case-by-case basis waiting for remixes to go viral to receive 1 million streams, or whether they should just bite the bullet and allow a broader license to regenerate content on certain platforms provided they receive attribution and revenue.
This pressure is exacerbated to the extent that an overly strict environment incentivises rights-washed uncanny valley knock-offs. The power of the rightsholder in these instances is emboldened only to the extent that prosumers want to maintain protected elements in the regenerated work. This will of course often be the case. For example, the techno Taylor Swift knock off needs the protected Taylor Swift material in its genre swapped version. Yet, where an individual wants to regenerate an underground artist with a unique style in the hopes of creating something similar, restricting access may have the opposite effect. Although maintaining some of the protected elements in the regenerated version may have been preferred, they will be substituted out for alternatives if doing so will allow its dissemination.
In turn, artists with strong senses of persona whose regenerated works are more likely to form around certain tentpoles within the music (such as famous choruses or imagery) will have greater power than those whose expression is most evident in attributes that are not copyrightable (such as their specific style). Artists who compose the most memetic elements, authored to invite endless remix in an array of different styles and genres, may find the most viral success in platform architectures in the era of interactive media, maintaining artist visibility across regeneration.
5. Conclusion
When we consider all of the above, we are faced with the prospective dilemma of copyright’s relationship with interactive media economics. Copyright’s goal is both to incentivise innovative production and reward authors for their work. When faced with this new form of production, copyright appears to drive it directly towards large prosumer social media platforms, given they may be the only spaces that can legally support the mass of copyrights necessary for widespread immediate media regeneration. Yet, the more powerful these platforms are, the less power rightsholders have to determine the terms by which their works are enjoyed within these platforms.
The worse the terms for rightsholders, the less value in the consumption of the works on-platform, placing greater value on attribution and visibility in hopes of receiving an alternative revenue. Yet, if works can easily be regenerated into similar shareable versions, strict copyright enforcement may incentivise creation of rights-washed public domain versions which are more amenable to new forms of interactivity, further pressuring rightsholders to allow remixing of their works to maintain attribution.
Thus, copyright’s friction with new media may embolden a market dynamic that (in many instances) undermines its own value for rightsholders, reliant on control mechanisms in an era where uncontrolled spread and infinite remix is paramount. Those who provide the tools and the space for this new form of prosumption may then reap the spoils. Greater content fluidity thus throws many spanners in the works of legacy comprehensions around copyright market incentives that demands nuanced practical analysis, as we enter an era of foundationally different creative consumption, production and dissemination.
Citation: Zachary Cooper, Dams For The Infinite River: Limits To Copyright’s Power Over The Next-Generation of Generative AI Media, The Law & Technology & Economics of AI (ed. Adrian Kuenzler, Thibault Schrepel & Volker Stocker), Network Law Review, Summer 2025.
References:
- [1] This is enshrined in the US Constitution, Article I, Section 8, Clause 8. European copyright legislation, despite its greater foundations in droits d’auteurs, still primarily justifies itself in its Recitals as on the basis of stimulating a single European market. See for example: Directive 2001/29/EC of the European Parliament and of the Council of 22 May 2001 on the harmonisation of certain aspects of copyright and related rights in the information society, OJ L 167 Recitals 1-4.
- [2] We may consider the ongoing 2025 copyright dispute over Anne Frank’s diary and wonder whether the private thoughts of a child truly required economic incentive at the time, let alone to this day. See: Iris Toepoel & Etienne Valk, Geoblocking measures sufficient to prevent a “communication to the public”? The CJEU gets a second chance, Kluwer Copyright Blog (2024), available at: https://copyrightblog.kluweriplaw.com/2024/10/31/geoblocking-measures-sufficient-to-prevent-a-communication-to-the-public-the-cjeu-gets-a-second-chance/; Mark Lemley, Faith-Based Intellectual Property, 62 UCLA L. REV. 1328 (2015)
- [3] Zachary Cooper, The AI Authorship Distraction: Why Copyright Should Not Be Dichotomised Based on Generative AI Use, SSRN (2024).
- [4] See, for example: Andrew Tarantola, YouTube’s new AI music remixer could let you swap genres, Digital Trends (2024), available at: https://www.digitaltrends.com/computing/youtube-testing-ai-music-remixer-shorts/; Team Suno, Introducing v4.5, Suno Blog (2025), available at : https://suno.com/blog/introducing-v4-5
- [5] Jack Shepherd, 25 Essential TikTok Statistics You Need To Know in in 2025, Social Shepherd (2025), available at: https://thesocialshepherd.com/blog/tiktok-statistics
- [6] See, for example: Luminate, 2024 Year-End Music Report, Luminate (2025); Julia Stoll, Video streaming worldwide – statistics & facts, Statista (2024), available at: https://www.statista.com/topics/7527/video-streaming-worldwide/
- [7] Ari Herstand, Best Music Distribution Companies in 2025 – Full Comparison Chart, Ari’s Take (2025), available at: https://aristake.com/digital-distribution-comparison
- [8] Composer Jennifer Walshe writes of her friend’s daughter who streams Taylor Swift on mute when she listens to her vinyl to ensure it counts to her Spotify Wrapped. See: Jennifer Walshe, The Entanglement – AI, Platform Musicking and the Future of Music, The Journal of Music (2024), available at: https://journalofmusic.com/opinion/entanglement-ai-platform-musicking-and-future-music; Marek Poliks, [Superlecture]: Nobody Listens to Music Anymore (Marek), Disintegrator, Apple Podcasts (Dec. 30, 2024); IFPI, Global Music Report 2025, IFPI (2025); Glenn Peoples, Music Merch Is Big Business, But It Won’t Deliver Big Growth, Billboard (2024); Luminate, 2024 Year-End Music Report, Luminate (2025)
- [9] It would take an estimated 10 million streams at .003 cents a stream with a 50/50 label split to reach the US federal minimum wage. Thanks to Marek Polik for the idea to compare musicians with influencers. See: Daniel Tencer, 7 in 10 Musical Artists Dissatisfied With Streaming Music Payouts, Survey Finds, Music Business Worldwide (2024), available at:https://www.musicbusinessworldwide.com/7-in-10-musical-artists-dissatisfied-with-streaming-music-payouts-survey-finds/; There are millions of artist profiles on Spotify, yet only a small fraction are generating money. Shouldn’t a higher percentage of all artists on Spotify be making money? Aren’t there 10 million of them?, Loud and Clear by Spotify, available at: https://loudandclear.byspotify.com/faqs/there-are-millions-of-artist-profiles-on-spotify-yet-only-a-small-fraction-are-generating-money-shouldnt-a-higher-percentage-of-all-artists-on-spotify-be-making-money-arent-there-8-milli/; https://www.keller-advisory.com/creators-uncovered-insights-from-a-nationally-representative-study-of-uscreators; Poliks, id.
- [10] Hugh McIntyre, Spotify Paid Out $9 Billion In 2023–How Does TikTok Compare?, Forbes (2024)
- [11] Dawn Chmielewski and David Shepardson, Universal Music Group artists to return to TikTok after new licensing pact,
- Reuters (2024); Ben Sisario, Taylor Swift’s Music Returns to TikTok Ahead of New Album, The New York Times (2024); TikTok VS… Music? (w/ Kristin Robinson), Money 4 Nothing, Apple Podcasts (2024)
- [12] Lucy Maguire, The Business of Brat, Vogue Business (2024)
- [13] Liz Pelly, The Ghosts in the Machine, Harper’s Magazine (2025), available at: https://harpers.org/archive/2025/01/the-ghosts-in-the-machine-liz-pelly-spotify-musicians/
- [14] Jay Peters, Netflix is removing nearly all of its interactive titles, The Verge (2024); Andrew Webster, Netflix has a new plan for games, The Verge (2025)
- [15] Notably, this is all only possible to the extent that each of these platforms have functional GenAI tools attached to them that allow interactivity. If open source tools dramatically trail proprietary models, then the models themselves may also compete as prosumer platforms. To the extent that certain actors have access to higher quality proprietary creative tools, this can in turn impact the competition over the most lucrative interactive social mediascape (although large prosumer platforms would seemingly have a lot of content to train their own models with, if need be).
- [16] The overwhelming majority of protected content that I submitted was not flagged by the content ID mechanisms.
- [17] 17 U.S.C. § 512 (2024)
- [18] Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on Copyright and Related Rights in the Digital Single Market, art. 17, 2019 O.J. (L 130) 92.; Paul Keller, How Filters fail (to meet the requirements of the DSM Directive), COMMUNIA (2020), available at: https://communia-association.org/2020/06/12/filters-fail-meet-requirements-dsm-directive/
- [19] DSM Directive, art. 17, id.
- [20] Cooper, supra note 3, at 30.
- [21] Katharine Trendacosta, Unfiltered: How YouTube’s Content ID Discourages Fair Use and Dictates What We See Online, Electronic Frontier Foundation (2020), available at: https://www.eff.org/wp/unfiltered-how-youtubes-content-id-discourages-fair-use-and-dictates-what-we-see-online
- [22] YouTube Copyright Transparency Report, Google Transparency Report (2025), available at: https://transparencyreport.google.com/youtube-copyright/
