back to list

AI expertise for innovative DataHub Europe project

Platform enables safe and data protection-compliant use of AI in business and science

2024/10/23 by

The Technical University of Darmstadt and the Hessian Center for Artificial Intelligence hessian.AI are playing a key role in the development of DataHub Europe, a new platform for the secure use and provision of data for training AI models. This platform, which was presented by Schwarz Digits and Deutsche Bahn AG at the German government's 2024 “Digital-Gipfel”, helps companies to efficiently implement specific solutions using artificial intelligence (AI) in a manner that complies with data protection regulations.

“DataHub Europe will close a crucial gap in the EU by providing high-quality data for training AI,” said Dr Volker Wissing, Federal Minister for Digital and Transport. “An exciting ecosystem and a much-needed data platform are emerging here that will take the training of AI models for our industry to the next level.”

In addition to Schwarz Digits, Deutsche Bahn AG, TU Darmstadt and hessian.AI, leading technology partners such as Aleph Alpha and the German Research Center for Artificial Intelligence (DFKI) as well as media companies are involved in DataHub Europe. Together, the partners are creating the technical and scientific basis for further developing artificial intelligence in a sovereign, trustworthy and secure environment. Leading AI researchers from Darmstadt, who bring together the expertise of several research institutions, have been involved in the project from the outset. The scientists Dr Patrick Schramowski and Manuel Brack (hessian.AI/TU Darmstadt/DFKI) are leading the project with the support of the AI innovation laboratory at hessian.AI; Simon Schampijer is the project manager. The participation was initiated by Mira Mezini and Kristian Kersting, who are both professors at TU Darmstadt and principal investigators at hessian.AI.

Platform for developing trustworthy and compliant AI applications

From left to right: Tobias Kehl (hessian.AI, Startup Rising), Dr Patrick Schramowski (DFKI), Professor Kristian Kersting (TU/hessian.AI), TU President Professor Tanja Brühl, Jonas Andrulis (Aleph Alpha), Professor Antonio Krüger (DFKI) and Stefan Schaffer (DFKI).
From left to right: Tobias Kehl (hessian.AI, Startup Rising), Dr Patrick Schramowski (DFKI), Professor Kristian Kersting (TU/hessian.AI), TU President Professor Tanja Brühl, Jonas Andrulis (Aleph Alpha), Professor Antonio Krüger (DFKI) and Stefan Schaffer (DFKI).

A key feature of DataHub Europe is that the platform enables companies, public institutions and scientific partners to develop and implement AI models in strict compliance with laws, data protection and security requirements. As part of the feasibility study, the researchers evaluated the quality of the underlying data and models regarding competence assessment, language comprehension and breadth of knowledge. In addition, data processing tools developed by the Darmstadt experts as part of the Occiglot initiative were incorporated into the training of the models. The pre-training of the models was carried out on the AI supercomputer 'fortytwo' from hessian.AI.

This contribution is instrumental in ensuring that companies and scientific institutions can benefit from high-quality AI solutions provided by DataHub Europe. An example of how this can be used in practice is the 'AuditGPT' application, which was presented at the Digital Summit. This AI-based tool, piloted by Deutsche Bahn AG and the Schwarz Group, among others, is designed to make auditing work in companies more efficient and automated. With the help of AuditGPT, confidential audit reports can be created more systematically and the workload on employees in audit departments can be reduced.

“I am very pleased that TU Darmstadt and hessian.AI are partners in DataHub Europe and that they are contributing their expertise and infrastructure with great added value,” said TU President Tanja Brühl. “This initiative impressively demonstrates that by working together, private-sector and scientific stakeholders can make a decisive contribution to Europe's digital sovereignty. We can train trustworthy AI models with and for sensitive data that meet the regulatory requirements in the European Union and take into account Europe's cultural and linguistic diversity. In doing so, we are taking a major step for the future of AI made in Europe.”