The European Health Data Space: are your collaboration agreements future-proof?

The European Union has laid the foundations for a new system for health data. After years of political wrangling, the European Health Data Space (“EHDS”) Regulation has now entered into force. This creates a framework to improve access to electronic health data and facilitate its use for research, innovation and regulatory purposes across the EU. 

For life sciences companies, universities, NHS partners, CROs and technology transfer offices, this is not simply a regulatory change. It has the potential to change the commercial value of health data and the standard positions in many existing collaboration agreements.

Data is becoming a strategic asset

Traditionally, life sciences collaborations have tended to focus on intellectual property – specifically patents. Negotiations often centred on ownership of foreground IP, licensing rights, publication restrictions and revenue-sharing arrangements.

It is now clear that access to high-quality datasets is becoming important in its own right. The number of big tech companies entering the health data industry is a sure sign of that. The EHDS intends to enable the secure reuse of electronic health data for research and innovation purposes at scale across Europe, as well as giving individuals more control over their data. 

Organisations may find that data access rights negotiated several years ago no longer reflect the strategic importance of the underlying datasets. A collaboration that was primarily about generating patentable inventions could generate significant further value from the data it generates – e.g. data-driven insights, training AI or biomarker discovery.

Existing contractual assumptions may be challenged

Many collaboration agreements contain provisions that were drafted before the EHDS framework was finalised. Common contractual assumptions may now deserve closer scrutiny.

For example:

• Do data-use provisions allow sufficient flexibility for future secondary uses?
• Are data access rights limited to specific projects or fields of use?
• Who benefits if a dataset becomes significantly more valuable because it can be combined with other European health datasets?
• Have the parties adequately addressed rights in derived datasets, algorithms, models and outputs?
• Are governance structures capable of accommodating future regulatory requirements?

In many agreements, these issues receive considerably less attention than ownership of foreground intellectual property. Now may be a good time to re-assess that.

The AI dimension

The importance of data governance is further amplified by the rapid growth of AI in healthcare and life sciences.

Many organisations are exploring the use of machine learning tools for drug discovery, diagnostics, patient stratification and clinical decision support. The value of these tools is often heavily dependent on access to large, diverse and well-structured datasets.

Some organisations are realising the opportunities and risks associated with AI and large health datasets. This often takes the form of blanket bans on the use of AI, but a more nuanced focus on terms such as data access, permitted uses and audit rights is likely to serve the organisation better in the long run. Rather than prohibiting AI outright, collaboration agreements should address the specific contractual mechanisms that enable organisations to leverage data for AI purposes while managing risk. Three areas merit particular attention:

• Permitted use definitions. Traditional data-use clauses may authorise "research purposes" or use within a defined "field of use", but these formulations were rarely drafted with AI in mind. Organisations should consider whether their agreements expressly permit the use of shared datasets for training, validating and fine-tuning AI and machine learning models — and whether that permission extends to commercial deployment of the resulting tools, or is limited to the collaboration itself.
• Ownership of AI-generated outputs. Collaboration agreements typically address ownership of foreground intellectual property, but trained models, model weights and AI-generated predictions or insights often fall outside those traditional categories. Where one party contributes the data and the other contributes the AI capability, the question of who owns the resulting model — and who can exploit it — needs to be addressed expressly. Silence on this point is an invitation to dispute.
• Audit and explainability rights. As AI systems in healthcare increasingly fall within the scope of the EU AI Act's high-risk classification, contractual audit rights should go beyond traditional data protection compliance. Organisations may need the ability to inspect training data provenance, review model documentation and require a degree of algorithmic transparency — not only as a matter of good governance, but to meet emerging regulatory expectations. 

Organisations that treat data merely as the raw ingredient for research may be missing an opportunity – and could be creating risk.

Practical steps for organisations

The EHDS does not mean every collaboration agreement requires renegotiation. However, organisations will benefit from keeping an eye on the development of the EHDS and should think about how their template collaboration and licensing agreements treat data.

Key questions include:

1. Whether existing data rights are aligned with new strategic objectives.
2. Whether agreements clearly address rights in derived data and AI-generated outputs.
3. Whether governance mechanisms can accommodate evolving regulatory requirements.
4. Whether value-sharing arrangements appropriately reflect the contribution of data assets.
5. Whether future collaborations should place greater emphasis on data access alongside traditional IP provisions.

Looking ahead

The EHDS is part of a broader shift in how health data is viewed across Europe. While intellectual property will remain central to life sciences innovation, data is increasingly becoming a core commercial asset in its own right.

For organisations entering new collaborations, the question is no longer simply who owns the IP. It is also who can access, use and generate value from the data. Those who address that question proactively are likely to be better positioned to unlock opportunities arising from the next generation of data-driven life sciences innovation.