News
Artificial Intelligence (AI) in Scientific Research: key takeaways from the European Commission’s 2025 report
On 8 October 2025, the European Commission’s Joint Research Centre (JRC) published its report “The Role of Artificial Intelligence in Scientific Research” (AI in Science Report), which provides a foundation for future European policy on AI in science and aims to help build a trustworthy, responsible, and transparent use of AI in research.
Key legal and ethical considerations: The JRC highlights several main issues that Life Science organisations must consider when using AI.
👉 Official report available here: European Commission JRC Report - “The Role of Artificial Intelligence in Scientific Research” (JRC143482)
1. Human oversight and accountability 🧍🏻♂️🧍🏾♀️
The AI in Science Report emphasises that AI should always remain a tool, not an author. Researchers are fully accountable for the integrity of their results, even when they use AI for literature reviews, data analysis, or writing assistance. This principle aligns with the European Code of Conduct for Research Integrity, and the Ethics Guidelines for Trustworthy AI, both of which underscore the importance of human agency and control.
For Life Science organisations, this means implementing clear human oversight at every stage of the research cycle. AI-generated findings should be verified and documented, with teams trained to recognise and mitigate the limitations and potential biases of AI models. Developing AI literacy is now essential to maintaining both scientific integrity and regulatory compliance.
2. Transparency and explainability 🔎
Transparency is essential for public trust and scientific credibility. The AI in Science Report calls for researchers to disclose when and how AI tools are used, especially if they significantly influence outcomes. Since most of the AI systems (i.e. generative or deep learning models) function as “black boxes”, explainability becomes essential for reproducibility and peer review.
In practice, organisations should maintain records of AI use, documenting model types, datasets, and parameters, and provide clarity about the uncertainty or variability in AI outputs. These practices also support compliance with the EU General Data Protection Regulation (GDPR), which requires transparency and accountability when processing personal data.
- Numéroter « 3. Bias and fairness » et « 4. Privacy, data protection, and dual-use risks »
AI systems are only as fair as the data they are trained on. If datasets underrepresent certain populations or research contexts, results may be skewed. In biomedical research, for example, biased datasets can lead to models that perform worse for underrepresented groups or produce incomplete findings.
The JRC stresses the need to assess datasets and models for potential bias and to implement fairness checks throughout development. In the Life Sciences, this might involve ensuring diversity in clinical data, reviewing model assumptions, and including independent ethical review of AI-driven research designs.
- Privacy, data protection, and dual-use risks 🔐
The AI in Science Report reinforces that responsible AI requires strong data governance and strict compliance with data protection law. Any AI use involving personal data must align with the GDPR. Researchers must establish a clear legal basis for personal data processing, ensure data minimisation, and apply safeguards such as pseudonymisation. When dealing with health and genomic data, organisations should also secure ethical approval and implement rigorous security safeguards.
In addition, the AI in Science Report draws attention to dual-use risks, where research designed for beneficial purposes could be repurposed for harm. For example, AI models developed for protein modelling might also enable the design of hazardous biological materials. The JRC urges researchers to assess such risks early and to integrate biosafety and ethical risk assessments into their AI governance frameworks.
5. The EU Artificial Intelligence Act (AI Act) and the research exemption 🤖
The EU AI Act creates a risk-based framework for AI governance. While the EU AI Act includes a “research exemption” (Article 2(6)) that excludes most scientific research from its rules, this protection ends once research outputs are deployed or commercialised.
For Life Science organisations, this means that an AI system used for internal research may later fall within the scope of the EU AI Act if it is developed into a diagnostic tool or marketed solution. To manage this transition, the JRC recommends adopting compliance-by-design and ethics-by-design principles, embedding regulatory and ethical considerations from the outset of the project.
Conclusion and practical steps ✅
The AI in Science Report confirms that responsible AI is a matter of both scientific quality and legal compliance. For organisations in the Life Sciences, implementing structured governance for AI-driven research is no longer optional - it is an operational necessity that supports credibility, mitigates risk, and ensures readiness for future regulatory scrutiny.
Practical measures to consider include:
By embedding these measures, Life Science organisations can demonstrate compliance, uphold ethical standards, and position themselves as leaders in responsible innovation. As the JRC rightly concludes, the future of research will rely on both technological advancement and principled governance.
📣 Final note
If you’re using AI tools in your Life Science programs and would like to understand more about your responsibilities, data governance, and GDPR compliance measures, contact MyData-TRUST. Our team of privacy experts would be happy to discuss your specific situation, identify a tailored solution, and support you in its implementation.
