The event focused on the Evidence Review Report and Scientific Opinion delivered by the Scientific Advice Mechanism (SAM) to Ekaterina Zaharieva, European Commissioner for Startups, Research and Innovation, in April. The Advisors’ recommendations will inform the European Commission’s proposed Advanced Materials Act, expected by the end of 2026. Building on an earlier webinar on strategic autonomy, this discussion focused on how digitisation, simulation and AI are reshaping materials research and innovation.
Setting the scene for the webinar, Geoffroy Delamare of the European Commission's SAM Secretariat outlined the work undertaken by the Scientific Advice Mechanism. He highlighted the two key questions posed by Commissioner Zaharieva: how advanced materials can support the EU's strategic autonomy, and how cross-fertilisation of innovation in advanced materials can be enhanced. AI and digitisation are central to both challenges.
Anke Weidenkaff, Co-Chair of the SAPEA Working Group, presented the main findings of the Evidence Review Report. She emphasised the need for circular strategies to reduce dependence on raw materials, long-term funding for fundamental research, and stronger coordination across Europe’s advanced materials ecosystem. Highlighting the Report’s focus on digitisation, simulation and AI, she described these technologies as key enablers of sustainability and circularity. She concluded that materials research is moving away from traditional, hypothesis-led approaches towards a more data-driven and ultimately ‘AI-native’ model of discovery.
Representing the Group of Chief Scientific Advisors, Dimitra Simeonidou outlined the main recommendations of the Scientific Opinion. The overarching aim, she explained, is to support Europe’s strategic autonomy, industrial competitiveness and economic growth while maintaining high standards of safety and sustainability. Among the challenges identified were fragmented data systems and a lack of interoperability. Recommendations included creating enhanced FAIR (Findable, Accessible, Interoperable and Reusable) data spaces that incorporate safety and sustainability information, achieving common data standards, and unlocking the scientific literature. The Opinion also recommends enhanced use of modelling, open experimentation, digital twins, self-driving laboratories and AI across the entire value chain.
Nicola Marzari, a member of the SAPEA Working Group, described a “materials revolution” driven by the convergence of simulation, data and AI. He pointed to rapid advances in quantum-mechanical simulation, significant investment by large technology companies and start-ups, and Europe’s strong position in first-principles simulation and materials databases. Machine learning, large language models and AI agents, he argued, are creating a paradigm shift in how materials research is conducted.
Karsten Reuter of the Fritz Haber Institute of the Max Planck Society explored the practical implications of these developments for experimental science. He showed how automation, robotics and AI are combining to create “self-driving laboratories” capable of increasingly autonomous research. Using a catalyst-screening project with Siemens Energy and BASF as an example, he explained how AI-guided systems can rapidly identify promising materials, dramatically shortening a process that previously took years.
The Q&A session, chaired by Euro-CASE Secretary General Patrick Maestro, covered a wide range of topics.
Discussing innovation, Nicola Marzari argued that Europe is highly innovative yet often struggles to translate ideas into commercial technologies. He described AI as a disruptive technology with the potential to create a more level playing field. Karsten Reuter agreed, noting that AI is already attracting significant investment and stimulating the growth of new start-ups.
On data governance and sovereignty, Dimitra Simeonidou highlighted the importance of open data as a driver of innovation, underlining the need for data that is reusable, traceable and trustworthy. She stressed that data access can be enabled through the European ecosystem, while remaining mindful of geopolitical and national security considerations. Marzari addressed industry's reluctance to share data, suggesting that sharing models rather than datasets could offer a practical way forward.
The discussion also explored the challenge of AI hallucinations. Reuter suggested that improving access to high-quality data is reducing the problem. He also pointed to extensive research into AI reasoning and understanding. Simeonidou agreed that hallucinations are becoming less common as larger volumes of data become available. She also highlighted the importance of explainability and noted that open-source environments can support greater transparency.
When asked about start-ups and the protection of commercially sensitive information, Marzari proposed an “app store” model as a potential solution, in which bespoke capabilities are built on top of common platforms.
Addressing concerns about errors and biases in density functional theory (DFT) training data, Marzari recommended the use of robust verification approaches. Looking ahead, Reuter described the potential of self-driving labs that integrate simulation and experimental data, allowing the two approaches to validate one another. He also highlighted the ongoing challenge of ensuring reproducibility across different laboratories.
Another theme was how to accelerate the journey from laboratory research to market deployment. Weidenkaff discussed the challenge of incorporating new knowledge from chemistry and the opportunities to combine real-world and training data. Reuter suggested that synthesis methods are a critical area for further development, identifying the automated synthesis of inorganic materials, supported by AI, as a promising field of research.
The panel also considered AI’s capacity to generate genuinely novel insights. Marzari argued that AI can free researchers from routine tasks, giving them more time to pursue creative and challenging scientific questions. Reuter suggested that AI’s role could be further transformed by developing open-ended systems capable of reasoning. In the meantime, AI is already enabling researchers to focus more of their efforts on innovative solutions.
On education and skills, Reuter called for fundamental changes to degree programmes to reflect AI's growing importance in research. He also emphasised the need to retrain established researchers so they can use AI-assisted methods effectively throughout their careers. In terms of future skills, Marzari emphasised that strong domain expertise remains essential, with AI best suited to technical and routine tasks. Reuter added that the rapid pace of technological change makes it impossible to predict the skills scientists will require throughout their careers. However, he argued that a solid grounding in disciplines such as physics, chemistry and materials science will remain fundamental, complemented by a working knowledge of programming and AI.
Addressing the potential of agentic AI systems, Nicola Marzari cautioned that the technology remains in its early stages and should undergo rigorous benchmarking before researchers can place full confidence in its outputs. Karsten Reuter agreed, noting that for now scientists continue to play a central role in setting research questions, designing experiments and overseeing automated labs. It is ultimately researchers who work with AI to interpret results and translate them into scientific knowledge and insight.
Asked whether self-driving labs can generate genuinely novel scientific discoveries, Marzari highlighted Europe’s unique opportunity to lead by adopting common standards and improving interoperability, thereby unlocking new opportunities for innovation.
