Blog post by Stephen Wyber
Within a few minutes of each other yesterday in Strasbourg, Members of the European Parliament voted through two motions relevant to how we regulate knowledge creation, sharing and application to support research and innovation. Neither has any immediate legal impact, but still set a direction that may well shape future, legislative, measures, for better or for worse.
And that direction is… confused.
The first to be voted on was a report on Generative AI and Copyright, prepared on the initiative of the Parliament’s Legal Affairs Committee, and led by Axel Voss MEP. The second was a motion for resolution on the upcoming European Research Area Act, led by Eszter Lakos MEP.
This blog explores what each says about how the EU should regulate knowledge, and where the risk of confusion arises.
The Texts
Axel Voss’ Report comes from a basic position of trying to find a response to what the copyright industries claim to be the problems created by generative AI. Both the original and the final versions of the report are clear that while they are supportive of innovation, they put the interests of copyright holders first.
Among its more detailed recommendations, the report focuses on how to accelerate the creation of licensing markets for AI training, notably by granting rightholders extensive powers to force AI actors into deals. It goes particularly far when it comes to news media, suggesting that they should have ‘full control’ over the use of their content for training purposes and any activity in potentially competing fields.
The report urges haste, calling for a short-term review of the implementation of existing laws, for example, to impose much greater transparency obligations, as well as to oblige licensing payments under a rebuttable presumption that copyright works have been used, even before any law to that effect is passed.
Eszter Lakos’s Motion runs through the different themes on the table in the context of the European Research Area Act, included in the mission letter given to Research Commissioner Zaharieva as the core measure to achieve the Fifth Freedom – the freedom of circulation of knowledge and researchers.
The same themes that have appeared in the Commission’s own proposals and updates are all there. There are calls for measures to increase investment in R&D, to support coordination of spending, and to uphold freedom of scientific research, although the motion also suggests that this needs its own parallel legislative initiative. It finally includes references to support for researcher mobility, as well as the open sharing of knowledge.
Opposing Motions?
There is already an obvious tension between the Voss text’s explicit decision to put innovation in second place, and the Lakos text’s strong support for more collaboration with industry.
While the Voss text does include an apparently helpful paragraph protecting research against hurdles, this is undermined by its limitation to non-commercial research. Anything with potential commercial applications, however uncertain, would still need to seek extensive authorisations that would likely be expensive or simply unforthcoming. This would likely make it next to impossible for researchers involved in industry collaborations, public-private partnerships, or spin offs to benefit.
Meanwhile, the Lakos motion seeks greater private-sector involvement in the R&I ecosystem, as well as a better uptake of research-originated ideas by industry. It is clear that the goal from that perspective is to remove barriers between non-commercial and commercial research, not to add them.
Admittedly, the Voss report is perhaps more nuanced than the impression given above, though not in a way that brings clarity. It bears the marks of disagreement between different creative sector industries about their own approach to AI. Alongside the focus on mandating licensing to secure revenues (the report at one point suggests that use of copyrighted material could even be made mandatory for AI trainers), it also strongly emphasises the need to enable opt outs entirely from AI training (which risks reducing the overall power of datasets to support safe and high-performing models). The report covers other areas too, outside the scope of this blog.
The irony of the Voss report is that Axel Voss himself has championed much more helpful provisions for research elsewhere, such as in his amendments on the Digital Omnibus on AI (see amendments 111 and 112 in particular).
Two Europes
In short, within a few minutes, the European Parliament passed motions that both support Europe’s ability to advance research and innovate, and go against it. These illustrate the conflicting and siloed narratives that too often mark the EU’s approach.
The potential tragedy is that despite the noise, it is far from sure that there is a choice to be made between enabling the research and innovation sector to develop and apply AI, and Europe’s future creativity.
Yet this becomes inevitable with the heavily copyright-focused approach taken by the Voss report. This may work well for those who stand to gain most from (administering) new licensing markets, and for those best able to pay the fees they generate. But for researchers working in industry collaborations, the libraries that serve them, and the users who benefit from access to a wider range of tools and applications, it imposes costs without clear benefit. Other routes exist. They just do not appear to have been seriously considered.
As the EU moves from non-binding instruments to hard-law measures, it needs to choose: a coherent approach that takes account of all priorities and options, or more of the same contradictions. In particular, copyright is not always the solution. It should not be treated as if it were.
11 March 2026
