☀️ Summer Reading Picks (August 2025): AI, Copyright, Deepfakes and Media in Transition
This month’s reading list underscores the messy but fascinating intersections of AI, copyright, and media economics. Taken together, these five pieces show how policy choices, legal experiments, and audience behaviors are colliding to reshape creative and information ecosystems.
As with last month, I encourage you to dive into the full texts themselves. My summaries are just an entry point into some of the most important debates unfolding this summer.
The Policy Foundation: Who Actually Benefits from Licensing Regimes?
Julia Willemyns and Pedro Serôdio, “Who Actually Benefits from an AI Licensing Regime?”, Centre for British Progress
This report delivers a sobering verdict: despite the UK’s talent base, no general-purpose LLMs have been trained in the country. The reason, according to the authors, lies in its restrictive copyright framework. By forcing would-be developers into labyrinthine permissions for text and data mining (TDM), the UK effectively exports innovation elsewhere.
The core point is elegantly simple: a licensing regime that was meant to balance interests has instead created a market deterrent, with companies training abroad and only selling services back to the UK.
What I find compelling is the clarity with which the report links law to missed opportunity. Rather than stimulating remuneration for creators, the licensing approach risks hollowing out domestic innovation capacity.
The Identity Challenge: Deepfake Bills in Denmark and the Netherlands
P. Bernt Hugenholtz, “Deepfake Bills in Denmark and the Netherlands: Right Idea, Wrong Legal Framework”, Wolters Kluwer Copyright Blog
Here we encounter the frontier of AI identity misuse. Denmark and the Netherlands are considering legislation that would give people exclusive rights over their likenesses in AI-generated deepfakes, with protections lasting up to 50 years beyond death.
At first glance, this looks like a robust defence of dignity. But the critique here is that the proposed frameworks rely on copyright-style neighbouring rights, which may be poorly suited to tackle the fast-moving and borderless nature of AI-driven manipulation.
What resonates with me is the uneasy trade-off: these bills acknowledge real harms, but in doing so may create rigid rights structures that complicate enforcement while failing to address the underlying technological dynamics.
The Legal Reality: Inputs, Outputs, and Infringement Revisited
Andrés Guadamuz, “Revisiting Copyright Infringement in AI Inputs and Outputs,” TechnoLlama
Guadamuz, a reliable guide through these thickets, takes us into the grey zones of copyright law. His central observation: even if inputs are lawfully obtained, outputs can still infringe—depending on how the model generates them.
This is not a call for alarmism but for nuance. He demonstrates how traditional categories of copying, transformation, and originality strain under AI’s probabilistic methods.
What strikes me is his insistence that debates cannot be settled by input licensing alone. The question of how outputs interact with existing works is where the next wave of disputes will arise.
The Rights Framework: Opt-Outs, Opt-Ins, and Policy Design
Martin Kretschmer, Bartolomeo Meletti, Lionel Bently, Gabriele Cifrodelli, Magali Eben, Kristofer Erickson, Aline Iramina, Zihao Li, Luke Mcdonagh, Emma Perot, Luis Porangaba, Amy Thomas, Copyright and AI in the UK: Opting-In or Opting-Out?, GRUR International, 2025;, ikaf093, https://doi.org/10.1093/grurint/ikaf093
This recent academic piece evaluates the UK’s proposed TDM exception that allows rights-holders to reserve their rights—in other words, an opt-out regime. On paper, this aligns with European approaches. In practice, the authors argue, it creates significant barriers: implementing and policing opt-outs is technically burdensome and risks privileging incumbents over new entrants.
The article’s deeper value lies in showing how policy design details matter. A system that sounds like compromise can end up stifling both innovation and equitable access.
For me, it reinforces a broader lesson: in the AI era, the architecture of legal frameworks can be as decisive as the rights they proclaim.
The Market Reality: How Audiences Actually Find Content
Charlotte Tobitt, “Publisher Traffic Sources 2019–2025”, Press Gazette
Finally, we zoom out to the media ecosystem. Chartbeat data from over 500 publishers shows that Google remains dominant (96% of search referrals), while social referrals have eroded to under 13%. Meanwhile, external referrals have nearly doubled, and Reddit has emerged as a surprise growth story, with traffic up 220% since 2019—surpassing even Instagram.
This is a reminder that while legal and policy debates unfold, the practical economics of attention are shifting under our feet. Publishers and creators alike will need to recalibrate strategies around search dependence and emerging platforms.
The Larger Pattern: What These Pieces Reveal
Taken together, these readings highlight a familiar pattern: well-intentioned frameworks often misfire. Licensing regimes aimed at fairness can drive activity abroad. Deepfake protections risk entrenching unworkable rights. Input-focused copyright debates miss the output frontier. Opt-out systems introduce complexity rather than clarity. And even as laws evolve, market realities are being reshaped by attention flows outside the courtroom.
The lesson is that we are not just tinkering at the edges of copyright or media regulation—we are confronting a deeper challenge of how human institutions adapt to technologies that scale differently, copy differently, and compete differently.
My Key Takeaways
What stands out across these works is a shared commitment to clarity over certainty. Each author resists the temptation to offer silver bullets. Instead, they map terrain, expose contradictions, and urge humility.
For policymakers, practitioners, and readers alike, this is invaluable. It equips us not with ready-made solutions, but with sharper lenses to interpret a landscape in motion.
Written by Caroline De Cock, LL.M. , Head of Research.
