As artificial intelligence continues to shape society, the need for trustworthy AI has never been more pressing. Trustworthy AI is also increasingly tied to ethical and legal aspects of AI, as reflected in the European AI Act. TRUST-AI 2025 is a European workshop dedicated to fostering reflection and exchange on trustworthiness in AI, bringing together researchers and practitioners to discuss real-world applications, challenges, and methodologies. The workshop is organized as part of the European Conference on Artificial Intelligence 2025 ECAI 2025, in Bologna, Italy, on October 25-26, 2025. While frameworks for trustworthy AI are rapidly evolving, and trustworthiness requirements for AI are increasingly detailed in guidelines and legislation, it is necessary to establish common understanding and insight into how to realize trustworthy AI in specific business or sociotechnical contexts. In the workshop, we address trustworthy AI from a human-centred perspective over the lifetime of the AI system. The TRUST-AI workshop has seen substantial interest at ECAI 2025, with more than 150 conference participants expressing an interest in the workshop and 60 papers submitted. This interest indicates the current importance and relevance of research into trustworthy AI.
The TRUST-AI call for papers was openly distributed through the communication channels of the ECAI 2025 conference, the networks of the TRUST-AI organizers, and to relevant email lists and networks.
Authors were invited to make two types of submissions: short papers and position papers. The short paper category should include theoretical contributions, case experiences, or findings from empirical studies and has an indicative max. length of 9 pages excl. references. Position papers should be shorter contributions, presenting a specific position or open questions in need of reflection or discussion. Only accepted short papers are included in these proceedings. Accepted position papers are published on the workshop website and in a separate position paper report.
In total 60 papers were submitted to the workshop, of these 49 short papers. Following an initial editorial check for relevance, the papers were submitted to peer review. All papers submitted to peer review received feedback from three independent reviewers, organized in the Easychair editorial system. A single blind review process was applied. The review process for each paper was led by a lead reviewer from among the workshop organizers – each organizer leading the review process for 9-10 papers with one organizer taking the role of overseeing the process. Following peer review, decisions for each paper were made in an organizer review meeting.
Following review, papers could be accepted directly or conditional to revision. Seven short papers included in these proceedings were accepted directly, 15 were accepted following minor revision, and 15 following major revision. Revisions were checked by at least two organizers to assess adherence to reviewer feedback. All but two conditionally accepted papers were finally accepted following a first revision. The latter two were finally accepted following a second revision.
The program of TRUST-AI 2025 included five sessions for plenary presentations of short papers and two sessions for poster presentations of short and position papers. In addition, the program included a groupwork on future research challenges in the field of trustworthy AI.
The plenary sessions were structured according to five thematic sessions: (1) fairness, diversity, and bias-mitigation, (2) enhancing AI trustworthiness, (3) trust in socio-technical AI systems, (4) measurement and assessment of trustworthy AI, and (5) AI compliance and explainability.
For the purposes of these proceedings, the short papers presented in the two poster sessions are structured under the following three thematic headings: (6) enhancing performance and integrity, (7) reliability and integrity, and (8) philosophical and societal considerations.
Overall, the program covers the breadth of research within the field of trustworthy AI, and serves as a good basis for sharing and discussions during and following the workshop. As such, the workshop proceedings contributes to advancing this important research field.
The organization of TRUST-AI 2025 is based in a Horizon Europe research and innovation action THEMIS 5.0 – Human-centered Trustworthiness Optimisation in Hybrid Decision Support. The workshop organizers represent seven technology research organizations, all contributing to advancing the state of the art on technology and knowledge in support of trustworthy AI. The organizers’ responsibilities included the chairing of the workshop, the peer review process, and these proceedings. The organizers were: • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •
Ibomoiye Domor Mienye, University of Johannesburg, South Africa Ilaria Vascotto, University of Trieste, Italy Ilias Driouich, AMADEUS France, France Ilias Kanellos, ATC – Athens Technology Center, Greece Jacopo Morelli, University of Modena-Reggio Emilia, Italy Jerome Guzzi, IDSIA (USI-SUPSI), Switzerland Joel Bjervig, SINTEF, Norway Jules Soria, CEA, France Koffi Ismael Ouattara, Huawei MRC, France Laurens Devos, KU Leuven, Belgium Lennard Helmer, Fraunhofer IAIS, Germany Leonie Louisa Etzold, German Aerospace Center, Germany Mahboubehsadat Jazayeri, University of Milan, Italy Manuel Fernando Silva Rodrigues, University of Minho, Portugal María J. Villalobos Quesada, National eHealth Living Lab, Leiden University Medical Centre, Netherlands Marina Tropmann-Frick, Hamburg University of Applied Sciences, Germany Martin Krutský, Czech Technical University, Czech Rep.
Michael Farayola, Dublin City University, Ireland Nicola Bena, Università degli Studi di Milano, Italy Nikolaos Antonios Grammatikos, University of Pireus, Greece Paolo Ceravolo, University of Milan, Italy Paolo Falcarin, Ca’ Foscari University of Venice, Italy Priyabanta Sandulu, DFKI, Germany Purbasha Chowdhury, Ca’ Foscari University of Venice, Italy Pål Vegard Johnsen, SINTEF, Norway Richard Oliver Lane, QinetiQ, UK Samuel Hill, DFKI, Germany Thanasis Tsafonis, ATC – Athens Technology Center, Greece Tim Barz-Cech, HMS Analytical Software, Germany Tim Kosack, German Aerospace Center, Germany
Yiqing Zhao, Amazon, USA
The workshop was supported by EU Horizon Europe, HORIZON-CL4-2022-HUMAN-02-01, under the project THEMIS 5.0 (grant agreement No. 101121042).