Blog post by Stephen Wyber
The draft report of the European Parliament’s Legal Affairs Committee on generative AI and copyright, drafted by MEP Axel Voss, is now out. It represents a clear preference for a model where Europe is willing to sacrifice its ability to realise the potential of the technology to achieve research and wider competitiveness goals, in order to support the interests of major corporate rightholders.
Knowledge Rights 21 approaches questions around AI from the perspective that artificial intelligence, including generative artificial intelligence, has significant potential to advance research goals. Moreover, with much research today taking place through collaborations between research centres and hospitals, civil society and industry.
Such collaborations – often encouraged by governments themselves – enable researchers to accomplish tasks that they would not be able to do alone, and contribute to the broader effort to overcome the European ‘Innovation Paradox’.
The challenge is that while the EU’s research and innovation policies have long recognised this potential, copyright law has not. Its clumsy distinction between commercial and non-commercial uses (a concept that is less than 25 years old in the European legal context) has carried through into the way that the EU regulates text-and-data mining, and so (generative) AI training.
The ‘own initiative’ report published at the end of last week, as its name suggests, is the result by some – and in particular its promotor, Axel Voss, to respond to the loud concerns expressed by many in the cultural industries to ‘do something’ about AI. While such reports do not necessarily lead to legislative change, they have a certain ‘soft’ influence, and the draft is explicitly directed to the upcoming review of the EU’s Copyright in the Digital Single Market (DSM) Directive. It is, therefore, worth getting things right.
Broad Where It Should Be Narrow, Narrow Where It Should Be Broad
The draft could have taken the path of exploring, in a comprehensive way, the different options open for ensuring that artists and other creators can continue to benefit from the fruits of their work in the face of just the latest technological change. This would have had the merit of evaluating all potential tools against each other, as well as safeguarding research from being collateral damage.
However, this is not the path taken by the draft report. It is broad where it should be narrow, looking at all generative AI rather than just that which could directly compete with the cultural industries. And it is narrow where it should be broad, by looking only at copyright as a potential tool.
The perspective taken is clear from the selective quotation of Article 27 the Universal Declaration on Human Rights on the first page (which neglects the rights to share in scientific advancement and its benefits), and the lack of reference to Article 179 of the Treaty on the Functioning of the European Union (where the EU commits to encouraging research in order to support competitiveness). Given this, the draft report’s claims to be supportive of competitiveness should be taken with a pinch of salt.
The main winners from the proposal, if it remains as it is and is integrated into European law, would be major corporate rightsholders and collective management organisations best placed to demand payments, and on the other side, those with the resources to make them.
Sacrificing Tomorrow for the Sake of Yesterday?
Losers would be researchers who would notably face both a dramatic shrinking of material available for training, and a chilling of research freedom, if carrying out work within the context of a collaboration, or even if just developing models that one day have the potential to be the basis of marketable products.
This is because of the heavy-handed requirement to share an exhaustive list of all materials used for training, or otherwise make a payment for all copyrighted content in existence. The choice is between prohibitive costs, or enabling surveillance of researchers who will already be working in line with protocols and codes of practice designed to ensure good conduct.
While such a model may still work for deep-pocketed mega-companies, this will not be the case for the research sector. This is particularly worrying when it comes to training involving research outputs, which have overwhelmingly been funded by the public sector.
Even beyond research taking place (currently or in future) in collaboration with others, the draft report also continues to spread the questionable argument that even Article 3 of the DSM Directive (which focuses on purely public research) does not cover generative AI training.
Such an interpretation seems designed to further restrict Europe’s ability to keep up in AI technology, sacrificing tomorrow for the sake of yesterday. This is, however, inevitable when maximalist copyright is used as a regulatory tool in a space like this.
Problems, Not Solutions?
In addition to the challenges for research set out above, the draft also includes a number of steps that would be of questionable legality. First of all, it attempts to subvert the principle of national treatment under the Berne Convention by suggesting EU law should apply when working with European content, regardless of where training takes place. The issue with this becomes clear if you imagine a non-European government trying to regulate how Europeans access and work with non-European content.
Secondly, it suggests that rather than waiting to establish any legal basis for the proposed actions, there should already be obligations to pay in place. These would of course also affect researchers for all of the reasons set out above. If the EU is willing to act in this manner, this undermines the point of lawmaking in the first place.
Finally, the report falls into the fallacy promoted by those making the case for maximalist copyright in general: that the only route to ‘legality’ is through an assault course of permissions and bureaucracy that can be resolved through payments. This measures effectiveness in terms of money flowing into the hands of (corporate) rightholders, not what is actually good for society as a whole. Such claims need to be vigorously rejected.
Still Space to Reframe?
So what can be done with the report? As highlighted above, an important step would be to provide clarification as to its scope. It should be explicit that it is not intended to regulate research, and that any recommendations it makes should not impede the work of researchers.
Similarly, it should not prejudge that copyright is the optimal solution; instead, it should look at the relative merits of other tools (such as the Smart Fund promoted in the UK) in achieving genuine public interest goals, as well as enforcing existing laws to ensure compensation when AI outputs replicate existing creative works.
More broadly, the draft report underlines the need for legislators to stop forgetting about research when legislating. If Europe wants to have a coherent and comprehensive approach to promoting research, as a driver of competitiveness and means of addressing challenges alike (including those related to cultural production), it needs to design laws and initiatives in this way.
Such an approach, focused on real policy goals, is far less likely to lead to a zero-sum game than one that centres on specific tools such as copyright.
11 July 2025
