Conference Agenda

Post-AI copyright is in flux. Its purpose is to stimulate innovation by granting exclusive monopolies to authors over their works under the rationale that they would not invest time and money creating those works otherwise. This rationale has now been foundationally undermined, as creators can minimally invest to produce fully-formed professional works with the help of Generative AI (GenAI). In turn, copyright would now counter-purposively incentivise individuals to mine creativity in order to own as much as possible.

To defend against this, copyright frameworks worldwide have rejected granting copyright to purely AI-generated works, while confusedly setting up ill-defined “human authorship” thresholds for AI-assisted works. Unfortunately, the thresholds to receive copyright are neither harmonised nor enforceable, without any means of auditing AI use. Instead, courts have been reliant on a “track or trust” approach, wherein they are either dependent on authors’ disclosure, or reliant on watermarking solutions that are both easy to remove and must surveil an artists’ entire workflow to be of any utility. Yet without robust means of auditing and enforcement, AI authorship thresholds are rendered dysfunctional – allowing copyright over gluts of low-cost AI-generated content, foundationally undermining crucial notions of substantial similarity between works. How can this be resolved?

In this paper, I reconceptualise the challenges of AI output and copyright away from unpracticable “human authorship” thresholds towards the defining questions of scale and of form. Importantly, it is the first contribution, informed by conversations with the designers of some of the world’s most crucial AI creative tools, to robustly analyse the impacts of copyright on AI’s most radical implication for creativity - the newfound instant interactivity it enables in media, as videos, songs and images can be regenerated instantaneously into bespoke new creations. If media fluidity may be the defining modality of 21^st century creative innovation, it is nonetheless exactly what copyright disallows – remix without permission. We therefore find ourselves in a mirror era to the illegal download era, as we can easily illegally produce regenerated and personalised content without permission.

I thus outline core incoherencies between traditional copyright logics and new forms of fluid AI-generated production, especially in relation to derivative works and substantial similarity – rendering models “infringement machines”. I then outline how public copyright frameworks’ incoherence with contemporary creative production is driving regulation to private actors, who seek to resolve these intrinsic tensions through unprecedented regulation at the point of production. I survey the current state of licensing deals between media platforms, streaming platforms, major rightsholders, and AI companies, who are rewriting the rules as to which creativity is and isn’t allowed, and what its accompanying economic models will be.

Critically, these private deals are transforming the application of copyright in determining which AI models are allowed to be trained in the first place, and in turn, are legally entrenching control of the creative tools themselves. The paper thus draws upon recent “dynamic effects” literature to exhibit how private industry decisions may be reverse engineered to resolve unclear public copyright questions around AI model-training in order to remove the capacity for future competition in nascent AI-creative industries.

I thus exhibit how copyright’s core doctrines, such as fair use, are ceding relevancy over new modes of creative production to the laws of what the tools themselves allow based on the private rulemaking of AI companies. In turn, entirely legal creative production can be both hampered or taxed, pre-emptively paying a fee to legacy rightsholders for perceived unproven algorithmic similarity. In doing so, I push back on recent legal literature and policy recommendations that support stricter algorithmic regulation at the point of production, without adequately considering its radical doctrinal shift.

I conclude that if copyright frameworks do not adapt to contemporary post-GenAI logics around near-infinite production scale and instant remix fluidity, they will progressively become inconsequential as regulatory forces over 21^st century creative production, relinquishing control to the private sphere at the expense of future creators’ ability to openly create new innovative works.

I thus suggest alternative policy interventions, such as sui generis formalities, that are not dependent on impossible-to-audit authorship thresholds in order to tackle scale of production. Drawing from pre-emption doctrine, I argue that protections are required to ensure that new modes of original and fair use AI-enabled creativity are not regulated anti-competitively at the point of production to ensure free and open access to the core modalities of 21^st century creativity.