ELLIOT project launches to build more responsible AI in Europe

The project brings together a broad network of research and industry partners to develop next-generation AI solutions that can operate reliably in complex and dynamic data environments. The project leverages the LUMI supercomputer to support the pretraining and fine-tuning of large-scale foundation models, as well as research into enhancing core model capabilities such as reasoning and generalization. These activities facilitate the development of responsible AI in Europe.

Launched in July 2025, the new EU research and innovation project ELLIOT (European Large Open Multi-Modal Foundation Models for Robust Generalization on Arbitrary Data Streams) brings together 30 research and industry partners from across Europe. The project aims to develop the next generation of open, trustworthy, and multi-modal foundation models that can effectively utilize diverse data types.

The goal is to create AI models that accelerate innovation in key sectors, such as media, Earth observation, autonomous systems, and public services, while also supporting Europe’s technological sovereignty and global competitiveness.

Towards an open and sustainable AI infrastructure

The core aim of ELLIOT is to develop multi-modal foundation models capable of processing and integrating diverse data types, such as text, images, video, sensor data, and satellite imagery, and applying that knowledge across a wide range of use cases. These generalist models will be built on open-source principles, ensuring they are freely accessible and adaptable for researchers and developers.

Unlike many earlier AI models designed for narrow applications, ELLIOT seeks to create more general-purpose and flexible AI systems that perform reliably even in noisy, real-world data environments. This approach supports Europe’s strategic goal of technological independence while ensuring that AI development remains open, ethical, and socially beneficial. It also provides a strong, transparent, and scalable foundation for European researchers and companies to build upon.

Foundation model pretraining and fine-tuning supported by the LUMI supercomputer

ELLIOT leverages Europe’s cutting-edge supercomputing infrastructure, including the LUMI supercomputer operated by CSC – IT Center for Science, to support the pretraining of large-scale foundation models. The project also explores techniques to enhance core model capabilities, such as reasoning and generalization, and investigates approaches to reduce model size for efficient deployment across downstream applications.

– ELLIOT is not just a technical project – it is a strategic investment in the future of European AI research. By supporting the training of early-career researchers and providing open resources to the broader research community, the project helps foster a more inclusive and responsible AI ecosystem. Leveraging the capabilities of the LUMI supercomputer, ELLIOT strengthens the development of responsible and open AI across Europe,  states Jesse Harrison, Development Manager and ELLIOT project lead at CSC.

Image: Adobe Stock