Trustworthy AI (TAI), as defined by the EU Independent High-Level Expert Group on Artificial Intelligence, is characterized by three core qualities: lawfulness, ethical alignment, and technical robustness. These qualities must be considered across the AI lifecycle, including development, deployment, and use.

Recent work highlights that the software engineering community increasingly needs TAI practices, while also revealing a gap in ethical AI use. In parallel, AI for software architecture is expanding rapidly, yet the community still lacks a consolidated perspective on how TAI principles translate into architecture-specific practices.

This session’s objectives are: (1) Identify and structure key concerns related to TAI use in software architecture, (2) Prioritize concerns and recommendations through community voting, and (3) establish a shared foundation for follow-up efforts.

The session will be organized as follows. Participants will choose a discussion group related to a TAI core quality. Each group will collaboratively produce a list of key concerns, actionable recommendations, and a note on why each issue is architecture-specific.

Group representatives will then present their results to all participants, and these will be consolidated into a shared structured board. Participants will vote on the most important TAI4SA concerns and recommendations. Next, facilitators will present empirical insights on practitioners’ perspectives regarding GenAI use in architectural decision-making, followed by a second round of voting.