Leveraging multimodal data to advance Generative Artificial Intelligence applicability in biomedical research (GenAI4EU)

Successful proposals will deliver results that ensure:
• Researchers (including clinical researchers) gain access to robust, trustworthy and ethical Generative AI models that effectively advance biomedical research towards predictive and personalised medicine.
• Researchers have the skills and knowledge to use Generative AI models to synthesise scientific literature and large-scale multimodal data, applying necessary precautions to generate new knowledge and enable breakthrough discoveries.
• The research community benefits from advanced methodologies to assess and validate Generative AI models, ensuring accuracy, transparency, traceability and explainability in biomedical research applications.

Proposals must meet all of the following scope criteria:
• Core research focus on enabling or contributing to the destination “Developing and using new tools, technologies and digital solutions for a healthy society”.
• Development or repurposing of ethical, trustworthy, robust Generative AI models directed at advancing predictive and personalised medicine.
• Use of large-scale, high-quality multimodal biomedical data (real and/or synthetic) with justification of data choice for model training and optimisation.
• Implementation of at least two distinct proof-of-concept use cases in different medical fields, demonstrating added scientific and potential clinical value relative to existing methods.
• Active engagement of potential end users in model lifecycle (design, adaptation, testing) and consideration of sustainability aspects.
• Creation or adaptation of rigorous evaluation methodologies addressing performance, robustness, alignment, bias/confounding, explainability, usability and ELSI aspects, building on established European guidelines (Trustworthy AI, Responsible Use of Generative AI, etc.).
• Full compliance with FAIR data principles and GDPR requirements, adopting best practices from European research infrastructures.
• High standards of transparency and openness: open-source code where feasible; detailed documentation of model assumptions, processes and data management.
• Demonstrable EU added value: involvement of EU and Associated Country industrial AI developers (including startups), fostering open strategic autonomy.
• Exploitation of synergies with other Horizon Europe/Digital Europe actions and data platforms (missions, partnerships, EU data spaces, EOSC, EBRAINS, etc.) without focusing on building new infrastructures.
• Inclusion of social sciences and humanities expertise to ensure research relevance, societal impact and responsible innovation.
• Optional but encouraged utilisation of AI factory resources; optional collaboration with JRC for regulatory science and pre-normative frameworks post-approval.
• Allocation of budget for cross-project networking, joint meetings and exchange of best practices; clinical studies detailed in annex if applicable.