Copyright, Creativity, and Generative AI: The Research Sector’s Missing Seat at the Policy Table

There is a longstanding tendency in policy debates to narrowly interpret “creative industries” as shorthand for music, film, and the performing arts. This tradition certainly seemed respected in the 4 June workshop and hearing organised in the European Parliament JURI Committee on generative AI and copyright.

That shorthand is not just misleading: it’s dangerous. It risks cementing laws and practices that privilege one subset of creativity while undermining another that is just as essential: scientific and research-driven innovation. From pharmacology to climate modeling, from public health breakthroughs to social science analyses, these domains operate at the beating heart of intellectual creativity. And yet, they are routinely overlooked in AI policy debates across the EU and UK.

It’s time to broaden the lens. And more importantly, it’s time to increase the seats at the policymaking table.

The Creative Sector: Beyond the Stage and Screen

It bears repeating: the creative sector is not just about galleries and Grammys. WIPO, the EU, the UK’s DCMS, and others all define creativity far more expansively. WIPO’s framing includes “creations of the mind, such as inventions; literary and artistic works; designs”—a scope that plainly encompasses scientific and technical publications that reflect the work of researchers, software development, and research outputs that generate copyrightable or patentable IP. 

In the UK, the DCMS has long included “IT, software and computer services,” “architecture,” and “design” as creative sub-sectors, drawing directly from John Howkins’ definition of the creative economy, which explicitly includes R&D. 

Even if you are not willing to recognise the inherent creativity that lies in scientific discovery, it is undeniable that researchers are writers too, and innovation produces IPR…and economic benefits. And yet, EU and UK AI/copyright policy still centers on entertainment’s concerns.

AI as Partner, Not Plagiarist: The Research Sector’s Relationship with AI

Where entertainment industry stakeholders often frame AI as a threat to human creativity and livelihoods, the science and research sectors view it differently: as a powerful partner in discovery. 

As the UK’s Economic and Social Research Council (ESRC), part of UK Research and Innovation (UKRI), notes, AI in research isn’t just about replacing tasks—it’s about enhancing them. It helps “discover unexpected connections” and extract value from vast troves of literature, slashing weeks of work into hours. That’s not hypothetical: we’ve already seen AI used to fast-track the development of COVID-19 vaccines, and predict protein folding.

AI holds the promise to accelerate breakthroughs by spotting patterns humans can’t, automating data sifting, and enabling a scale of hypothesis generation that’s simply not feasible by manual means. Examples include expediting drug discovery and development, enhancing medical diagnostics, advancing materials science, improving climate change modeling, and enabling more sophisticated analysis in fields like genomics and social sciences.

The sector’s concern lies not in controlling the outputs of AI in the marketplace, but in ensuring lawful and affordable access to inputs: research articles, datasets, and scientific literature that are often locked behind copyright barriers or restrictive licensing. It’s not about “deepfakes of Mozart”: it’s about training an AI to screen for early-stage cancer or optimize sustainable agriculture.

But this promise only holds if researchers can access the raw material—data, in some cases stemming from copyrighted texts—to fuel these systems. And here lies the problem.

TDM Isn’t a Side Show: It’s the Show

Text and Data Mining (TDM) is not some niche quirk—it is core to how researchers leverage AI. It enables researchers to extract patterns, generate insights, and build models by analyzing vast corpuses of text and data—including copyrighted scholarly literature.

Yet both EU and UK copyright regimes have hamstrung this process.

The EU’s DSM Directive provides a mandatory TDM exception for scientific research under Article 3, but Article 4 muddies the waters with opt-outs that rightholders can impose via technical means. Meanwhile, in the UK, the fight over whether to expand TDM exceptions—especially for commercial and quasi-commercial research—remains heated and unresolved.

The upshot? Scientists find themselves stuck in a policy purgatory. Neither free to use copyrighted materials at scale, nor confident in the legal boundaries of doing so, they tread cautiously or not at all. The result isn’t caution—it’s a missed opportunity.

If policymakers continue to treat AI-related TDM as merely an entertainment industry concern, they risk throttling the very innovation they claim to champion. Research institutions that have lawful access to content should be able to mine it, without uncertainty or contractual or technical impediments. Anything else is a digital version of buying a book and being told you can’t read it with glasses on.

One-Size-Fits-All Won’t Fit the Future

What’s crystal clear is that science and entertainment are different creative beasts. One produces cultural works for consumption. The other generates knowledge as a public good. Their timelines, their incentives, their ecosystems—none of these map neatly onto each other. Yet policymakers continue to design copyright rules as if the entire “creative sector” operates on the same business model.

The dangers of applying entertainment-driven copyright rules to research contexts are manifold:

• Legal Chill: Scientists unsure of their rights may avoid using AI at all.
• Innovation Loss: Potential breakthroughs are thwarted by the inability to train on comprehensive datasets.
• Public Harm: Publicly funded research, often aimed at solving collective challenges, is rendered inert by access barriers.
• Economic Displacement: The EU and the UK risks losing AI R&D investment to jurisdictions with clearer, more AI research-friendly rules.

This is folly. Worse, it’s harmful. AI systems used for non-substitutive, transformative purposes—like research—deserve distinct treatment. The EU and UK should heed that perspective. Otherwise, they risk erecting regulatory firewalls that keep their own scientific communities from building the next vaccine, discovering the next climate solution, or cracking the next code of human behavior.

Don’t Let Silence Equal Exclusion

The entertainment sector is loud. It has experience, infrastructure, and money to spare for lobbying. The research community? Not so much. That’s why its interests get steamrolled—not by ill will, but by inertia. But the consequences are real: legal uncertainty, stifled discovery, and research deserts in data-intensive domains.

This is particularly pressing in light of the ongoing discussions around the European Research Area (ERA) Policy Agenda, which aims to shape the future of research and innovation across the EU. The ERA Action 4 on research careers and Action 7 on knowledge valorisation intersect directly with the AI and copyright debate, yet the research sector’s concerns around data access and generative AI have been largely absent from the conversation. With the ERA plenary meeting taking place in Gdańsk on 12–13 June 2025, this is a timely opportunity for policymakers to broaden the agenda and bring research voices into the fold.

So let’s stop equating visibility with value. Let’s make space—legally, politically, and rhetorically—for the research sector to speak and be heard. Not as a footnote. Not as an afterthought. But as a full creative sector in its own right.

Because in the end, a copyright policy that only protects yesterday’s outputs will never build tomorrow’s breakthroughs..

Written by Caroline De Cock, LL.M. , Head of Research.