The conference expands theoretical and practical understanding of the vulnerabilities inherent to machine learning, explores the robustness of learning algorithms and systems, and aims to build a unified scientific community for trustworthy machine learning. Topics include novel attacks on and defences for machine learning, secure and safe machine learning in practice, verification of algorithms and systems, machine-learning system security, privacy, forensic analysis, fairness and interpretability, trustworthy machine learning in cybersecurity applications and trustworthy data curation, with research, systematization-of-knowledge and position papers. Four workshops (New AdvML-Frontiers, MIRROR, TrustTFM and WICE) take place on 3 May 2027 ahead of the main conference, which also has a competition track featuring LatentArena and Ravensport. The general chair is Giovanni Apruzzese and the program chairs are Fabio Pierazzi and Florian Tramèr.