AK Explainability, Transparency, and Safety (ExTraSafe)

The mission of the work group (AK) ExTraSafe (Explainability, Transparency, Safety) of the AI section (FBKI) of the German Computer Science Society (GI) is to jointly investigate ExTraSafe properties in artificial intelligence systems at the intersection of pure AI research, legal regulation, and human-AI interaction. The AK ExTraSafe also intends to participate in regulatory debates and engage in science communication to provide scientific and legal expertise for the public. The AK was founded in 2026 by the steering committee of the AI section.

The current speaker is Benjamin Paaßen (Uni Bielefeld). Please approach them for requests to the AK (e.g. requests to join). You can also subscribe to the AK's mailing list here: https://lists.gi.de/postorius/lists/ak-extrasafe.lists.gi.de/ 

Current members of the AK are:

• Thomas Bartz-Beielstein (TH Köln)
• Hanna Drimalla (Uni Bielefeld)
• Barbara Hammer (Uni Bielefeld)
• Daniel Neider (TU Dortmund)
• Michael Kohlhase (FAU Nürnberg-Erlangen)
• Thomas Kosch (HU Berlin)
• Maximilian Kroker (Independent Lawyer)
• Benjamin Paaßen (Uni Bielefeld)
• Florian Rabe (FAU Nürnberg-Erlangen)
• Julius Schöning (HS Osnabrück)
• Tim Schrills (Uni Lübeck)
• Alexander Steen (Uni Greifswald)
• Felix Weitkämper (German University of Digital Sciences)
• Katharina Weitz (Fraunhofer HHI)
• Daniel Winteler (Fraunhofer)

August 11, 14:00-17:45, 2026 in the first floor of the Mehrzweckhochaus, Bibliotheksstraße 5, 28359 Bremen, Germany as part of the KI2026.

With the increasing deployment of AI systems in application scenarios that profoundly impact humans, demands on the trustworthiness of AI systems also increase, as exemplified prominently in the EU guidelines for trustworthy AI and the EU AI act. As key dimensions of trustworthiness, explainability, transparency, and safety (ExTraSafe) are particularly prominent and extensive AI research has focused on developing novel methods to enhance ExTraSafe properties of AI models and systems. The goal of this workshop is to engage in transdisciplinary discussion between AI research, human-AI interaction, AI practitioners, and law. Further, the workshop functions as annual meeting of the newly introduced work group ExTraSafe of the AI section of the German Computer Science Society (GI - FBKI - AK ExTraSafe).

Call for Papers

Key topics of interest are (but are not limited to):

• Legal regulation of AI systems, in particular papers on the legal compliance of AI systems in certain application domains and design implications of legal regulations.

• Practitioners' perspectives on ExTraSafe properties of AI systems in high-stakes application domains, e.g. automotive industry, industrial automation, healthcare, education, or agriculture.

• Analysis of ExTraSafe properties across the entire system lifecycle, especially with respect to architecture considerations, algorithm design, and evaluation practices.

• Approaches to social explainable artificial intelligence (sXAI), taking into account heterogeneous explanatory needs, social roles and contexts, incremental multi-step explanations, and/or multimodality.

• Novel approaches to the evaluation of ExTraSafe properties, especially in user and field studies.

Submission Formats

We invite two types of submissions:

• Research papers (up to 8 pages + references in Springer LNCS layout ; shorter papers are explicitly welcome), which will receive 10 min presentation time + 5 min discussion time in the workshop
• Discussion entries (up to 5000 characters of description directly in easy chair), which will receive 5 min presentation (impulse) time + 10 min discussion time in the workshop

Submission will be possible via easychair: https://easychair.org/my/conference?conf=extrasafe26 

Proceedings

The proceedings are published as CEUR Volume 4263.

Important Dates

• Submission Deadline: May 15, 2026 May 22, 2026

• Notification of Acceptance: May 29, 2026 June 09, 2026

• Early-Bird Registration Deadline: June 08June 15, 2026

• Workshop Date: August 11, 14:00-17:45, 2026

Schedule

The schedule for the Workshop is:

• 14:00-14:05 Welcome by the organizers and introduction to the workshop topic and agenda
• 14:05-14:50 Keynote Hannah Ruschemeier (Public Law and Digital Law, University of Osnabrück)

Compliance, ethical aspects, and regulation

• 14:50-15:05 Lara Kuhlmann, Alexander van der Staay, Maximilian Nebel, Annika Zemke, Christian Janiesch and Michael Henke. Technical, Organizational and Ethical Aspects of Integrating Enterprise Large Laguage Models into Existing IS Infrastructure (Research Paper)
• 15:05-15:20 Julius Schöning and Paul Wachter. From Plausible to Verifiable: Hybrid Data as a Reference Standard for Explainable and Compliant AI (Reseach Paper)
• 15:20-15:35 Hyeon Tak Shin. AI Sycophancy and the Regulatory Blind Spot: Toward an Affirmative Disclosure Duty for AI Literacy (Discussion Entry)

25 min Break

Robustness & Trust

• 16:00-16:15 Marvin Schneider, Jörn Schneider, Holger Voos and Jürgen Kaiser. Robustness Verification of Traffic Sign Recognition: Practical Feasibility from a Safety Perspective (Research Paper)
• 16:15-16:30 Jiaao Li, Matthis Großkreutz, Tobias Peters, Ingrid Scharlau and Benjamin Paassen. LLM Inconsistency under Paraphrase Attacks (Research Paper)
• 16:30-16:45 Tobias M. Peters and Philip Porwol. Does it Matter How We Talk About AI - Linguistic Anthropomorphism and its Effect on Trust in AI Systems (Discussion Entry)

Explainability

• 16:45-17:00 Marc Halbrügge. A New Paradigm for the Empirical Evaluation of the Understandability of LIME Visualizations (Research Paper)
• 17:00-17:15 Vivien Lohmer and Josephine B. Fisher. The Conversational Core Job as a Local Context in Game Explanations (Research Paper)
• 17:15-17:30 Polina Galkin. If anyone builds it, everyone thrives: On theneed of explainability-driven development (Discussion Entry)
• 17:30-17:45 Fulya Horozal, Dimitrios Dardanis, Stergios Chairistanidis, Georgios Fatouros, Robert Stefan Fejer and Mathias Horner. XR5.0 Trustworthy AI Framework for Explainable Human-AI Collaboration in Industrial XR Environments (Discussion Entry)

Organizers

The workshop is organized by the AK ExTraSafe of the AI section of the German Computer Science Society (GI - FBKI - AK ExTraSafe). The organization committee is:

• Benjamin Paaßen (Bielefeld University; DFKI), bpaassen@techfak.uni-bielefeld.de (Main Contact Person)

• Vera Schmitt (TU Berlin)

• Thomas Kosch (Humboldt-Universität zu Berlin)

• Julius Schöning (Osnabrück University of Applied Sciences)

• Hanna Drimalla (Bielefeld University)

• Florian Rabe (University Erlangen Nuremberg)

• Tim Schrills (University of Lübeck)

• Daniel Neider (TU Dortmund University; Research Center Trustworthy Data Science and Security; Lamarr Institute for Machine Learning and Artificial Intelligence)

• Peter Fettke (German Research Center for Artificial Intelligence (DFKI); Saarland University)