Enhancing Trustworthy AI in Safety-Critical Sectors

As a PhD candidate in Artificial Intelligence at the University of Bologna and an active member of the TUPLES project, Matteo Francobaldi recently presented a remarkable work at a seminar organized by Brown University’s esteemed CRUNCH Group.

The research, co-authored with Prof. Michele Lombardi, aims to address a crucial challenge in AI: achieving formal guarantees for machine learning systems—a requirement for their adoption in regulated and safety-critical fields such as healthcare and automotive.
This innovative framework tackles a significant gap: while Machine Learning and Neural Networks have made leaps forward, providing formal guarantees for their behavior remains a challenging and essential task, especially in regulated, high-risk sectors like automotive and healthcare.
With the recent AI Act by the European Union underlining the need for compliance, SMLE offers a pathway for creating models that meet these standards by ensuring comprehensive property satisfaction at training time.

This research aims to advance the reliability of AI applications, thereby promoting its responsible adoption, especially in safety-critical areas. To dive deeper, access the full preprint on arXiv here.