Mission without definition no more? Towards a definition of scientific research in EU information law

The idea of scientific research as an activity that deserves specific treatment under EU Law has been around for over 30 years. In the 1992 Rental and Lending Directive (Article 10(d)), it was highlighted as an area where exceptions to exclusive rights might apply. However, for a topic that has been with us for so long in information law, there is little in the way of formal definition in EU legislation that can guide interpretation. 

This has become an increasingly pressing question, for example with the discussion of whether the activities of LAION count as scientific research. However, with the move by the European Commission to include a full definition in the Digital Omnibus on Data in a new Article 4(38) of the General Data Protection Regulation (GDPR), things are coming to a head, fast.

The potential significance of a definition is high, given that this would be a reference for future legislation, as well as a guide to the Courts in interpreting EU law. With a wide range of amendments on the table from across the political spectrum in the European Parliament, now is a good time to take a look at what already exists, and what might exist in future. This piece takes inspiration from João Pedro Quintais’ analysis of the definition of research institutions published last year.
 

Clues not Clarity

The concept of ‘scientific research’ appears in numerous places in EU law, starting at least as early as the 1992 Rental and Lending Directive. However, at no point is there a clear effort to define what it is. This is particularly striking in the case of the 2019 Open Data Directive, which includes an entire chapter on scientific research data, but no definition of scientific research.

The strongest hints we get are in the 1996 Database Directive, which states that scientific research ‘covers both the natural sciences and the human sciences’ (Recital 36), a point repeated in the 2019 Directive on Copyright in the Digital Single Market (Recital 12).

Another thread is the idea that scientific research is guided by ethical standards. This appears in the current (2016) version of the GDPR (Recital 33), as well as in Recital 26 of the 2022 Data Governance Act.

Elsewhere, we need to extrapolate. Recital 42 of the 2001 InfoSoc Directive talks about ‘non-commercial… scientific research purposes’, implying that research can also be commercial. The 2024 AI Act also provides some room for speculation, for example in its Recital 109 which excludes ‘the need for compliance for persons who develop or use models for non-professional or scientific research purposes’. The fact that ‘non-professional’ purposes are presented as an alternative to ‘scientific research’ purposes implies that scientific research is ‘professional’. 

Article 2(6) of the AI Act notes that the ‘Regulation does not apply to AI systems or AI models, including their output, specifically developed and put into service for the sole purpose of scientific research and development’. The separation of research and development could be taken to imply that research does not incorporate development.

There is a similar implication in the Data Governance Act, whose Recital 25 states that ‘scientific research purposes should be understood to include any type of research-related purpose regardless of the organisational or financial structure of the research institution in question, with the exception of research that is being conducted by an undertaking with the aim of developing, enhancing or optimising products or services’. While this does clarify that scientific research is not exclusively carried out in specific environments or with specific funding, it does appear to exclude commercial development activities, at least in the context of this Regulation. 
 

Contours and Contentions

EU information law as it currently stands therefore only offers pointers, and no clear definition of what should count as scientific research. This is the vacuum into which the proposed new Article 4(38) of the GDPR steps, in the context of the Digital Omnibus. 

22 different amendments to this proposal have been submitted, across the four different committees providing views. A list of the relevant amendments is available here for reference. Four amendments simply call for a deletion of the paragraph, leaving us with 18 from all but one of the European Parliament’s political groups, raising the following open questions. 

Research, or Research and Development?: echoing the point raised by the Data Governance and AI Acts, there is a question up front about whether a future definition of scientific research should include development activities. The Commission does so in its proposal, and eight amendments agree, from the liberal Renew Group to more right wing parties. Some from the European People’s Party propose adding in that ‘privately-funded research’ is covered for clarity.

However, other parties (both left and far right) remove this explicit reference. While in some cases this may be driven by a desire to avoid introducing new concepts, it appears that for the Socialists and Democrats group, the intention is indeed to exclude more applied activities. This conclusion is reinforced by an S&D amendment to the recitals, arguing that the processing of data for scientific research should be ‘functionally separated from commercial monetization, tracking, or profiling activities’. This point also connects to the idea that researchers should be autonomous, as also argued by Renew and Green amendments.

A Requirement for Rigour?: another set of proposals focus on how research itself should be carried out. Echoing the LAION decision, two S&D amendments set out that scientific research should be methodical and systematic, as does a Green one. Meanwhile, two EPP amendments propose that research should be ‘creative and systematic’, as well as pointing to the Frascati Manual as a reference point.

Another dimension is the need for research to be documented and traceable, as proposed by Agius Saliba, while Renew and Green amendments underline the need for ‘verifiable and transparent results’. A potentially more difficult angle is the idea that there should be a ‘genuine intention’ to produce such results – something that may be difficult to prove in future. 

What Goals?: The Commission’s text underlines that research should ‘contribute to existing scientific knowledge or apply existing knowledge in novel ways’, and ‘to the growth of society’s general knowledge and wellbeing’. This is a phrasing that is left intact by many amendments. Nonetheless, EPP proposals rewrite this aspect, arguing that scientific research involves either ‘experimental or theoretical work undertaken primarily to acquire new knowledge of the underlying foundations of phenomena and observable facts without any particular application or use in view, or through original investigation undertaken in order to acquire new knowledge directed primarily towards a specific, practical aim or objective’. In doing so, they also remove the idea of the need for a public interest focus. 

In this context, it is worth highlighting the treatment of commercial research. While this has frequently been excluded in the past (by limiting exceptions to non-commercial scientific research), the Commission’s proposal makes it explicit that commercial research can fall within the definition here. Interestingly, there appears to be broad consensus behind this, with only those opposing any special treatment for scientific research removing reference completely here.  

Ethical Obligations?: Recalling the original GDPR and Data Governance Act, the role of ethics does appear in the Commission proposal: scientific research must ‘adhere to ethical standards in the relevant research area’. Many follow this line, but a number of S&D amendments add in the need to respect human autonomy and consent.

Is AI included?: The Commission steers clear of calling out AI, but some MEPs are less cautious. Many are positive however, with Groothuis et al (Renew) adding that ‘the development, training, testing and validation of new technologies, such as AI systems, AI models and AI agents may be topics of scientific research where it meets the conditions of this point’. 

Others limit applicability to uses prior to placement on the market (Tobé et al, Dahl, Kokalari), while only those on the far left and right try explicitly to exclude it altogether.

Where Next?

The focus of this contribution  has been on ‘raw’ amendments. Not only do these need to be condensed down into one version (alongside all the other recitals and articles) that can secure a majority in the Parliament, but then the Parliament’s version of the Omnibus needs to be reconciled with the Council’s. It seems unlikely that in such a political dossier, research will be the ‘make or break’ issue, and so there may well be an element of unpredictability in the definition that is ultimately agreed, depending as much on political horse-trading as anything.

Nonetheless, in the proposals already being made it is possible to see the shape of future political – and legal – discussions around what can and cannot benefit from special treatment under different laws, not just data protection. Concretely, we may see more detail in the CJEU’s upcoming consideration of the LAION case, as well as in the European Research Area Act later this year, and the proposed copyright reform next. The inclusion of applied research, the autonomy of researchers, the legal role of professional ethics, and expectations around methods and documentation may all find their way into EU law in a way that could have far-reaching implications.