Trusted AI Group

We build rigorous formal foundations for trustworthy AI — combining formal verification, reactive synthesis, and machine learning to provide safety guarantees for autonomous systems in safety-critical domains.

The Trusted AI Group is led by Prof. Bettina Könighofer at the Graz University of Technology. Our research work spans runtime monitoring & enforcement, safe reinforcement learning, reactive synthesis, and formal specifications for AI.

Research  ·  Publications  ·  Team

Institute of Information Security
Graz University of Technology
Sandgasse 38a
8010 Graz, Austria
bettina.koenighofer@tugraz.at

News

Bettina is giving a talk at the 12<sup>th</sup> Women in AI Meetup
Jun 04, 2026

Bettina is giving a talk at the 12th Women in AI Meetup

Bettina will talk about bilateral approaches for trustworthy autonomous intelligence at the Women in AI Meetup #12 Styria Edition, co-organized with Bilateral AI and Women in AI Austria. The event takes place on June 17, 2026 at 6 PM at TU Graz, Inffeldgasse 25D.

Invited Lecture at ECOOP 2026 Academy: Safe Autonomous Agents via Shielding
Jun 03, 2026

Invited Lecture at ECOOP 2026 Academy: Safe Autonomous Agents via Shielding

Bettina will give an invited lecture at the ECOOP 2026 Academy on July 2 in Brussels, covering the shielding pipeline from world model learning to shield synthesis with quantitative safety guarantees.

FMCAD 2026: Record Number of Submissions!
May 12, 2026

FMCAD 2026: Record Number of Submissions!

Paper submission for FMCAD 2026 has closed, and we are thrilled to announce a record number of submissions, the most in the history of the conference! We can’t wait to see everyone in Graz this September.

Welcome to the Group, Matthias Grill!
May 01, 2026

Welcome to the Group, Matthias Grill!

Matthias Grill joins the Trusted AI Group as a PhD researcher. His work focuses on bridging world model learning with formal safety. Welcome, Matthias!

Paper Accepted at ICML 2026
Apr 30, 2026

Paper Accepted at ICML 2026

Our paper “Efficient and Safe Molecular Assembly via Reinforcement Learning and Constraint Solving” has been accepted at ICML 2026. Congrats to all authors!

Upcoming Talk: Formal Verification of Neural Networks: Guarantees Beyond Testing
Apr 16, 2026

Upcoming Talk: Formal Verification of Neural Networks: Guarantees Beyond Testing

Laura Nenzi (University of Trieste) visits on April 30 at 12:00 in CCG EG 002 to talk about what it actually takes to prove a neural network safety.

FMCAD 2026: Invited Speakers Announced
Apr 14, 2026

FMCAD 2026: Invited Speakers Announced

Our invited speakers for FMCAD 2026 are Laura Kovács (TU Wien), Clark Barrett (Stanford), Claire Xenia Wolf (YosysHQ), and Travis Hance (MPI-SWS). We are really looking forward to having them in Graz!

Welcome to the Group, Matthias Grilz!
Apr 01, 2026

Welcome to the Group, Matthias Grilz!

Matthias Grilz joins the Trusted AI Group as a PhD researcher. His work focuses on safe constrained reinforcement learning and the integration of normative reasoning with probabilistic model checking. Welcome, Matthias!

Welcome to the Group, Liam Plank!
Jan 01, 2026

Welcome to the Group, Liam Plank!

Liam Plank joins the Trusted AI Group as a PhD researcher. His work focuses on safety and ethics in reinforcement learning. Welcome, Liam!

Student's List Award for Logic and Computability
Dec 09, 2025

Student's List Award for Logic and Computability

Bettina Könighofer and Stefan Pranger’s course “Logic and Computability” received the Student’s List Award, presented by student representatives for the best courses in the CS/SEM/ICE programs based on student feedback. Congratulations!

We're Hosting FMCAD 2026 at TU Graz
Dec 09, 2025

We're Hosting FMCAD 2026 at TU Graz

We are proud to host FMCAD 2026 in Graz! Join us September 14 to 18 at the Schumpeter Laboratory for one of the premier venues for formal methods research.

Upcoming Talk: Polynomial-time minimizable automata for omega-regular languages
Dec 05, 2025

Upcoming Talk: Polynomial-time minimizable automata for omega-regular languages

Rüdiger Ehlers (TU Clausthal) visits on December 18 at 10:00 in the ISEC meeting room (IFEG074) to share recent progress on a long-standing open problem in automata theory: polynomial-time minimization for automata over infinite words.

Shields for Safe Reinforcement Learning Published in CACM
Oct 20, 2025

Shields for Safe Reinforcement Learning Published in CACM

Our article on shielded reinforcement learning is out in Communications of the ACM! Co-authored with Roderick Bloem, Nils Jansen, Sebastian Junges, and Stefan Pranger, it surveys the shielding approach for safe RL and serves as an entry point to the field where formal methods and AI meet.

RV 2025: Keynote Speakers Announced
Jun 17, 2025

RV 2025: Keynote Speakers Announced

We are delighted to announce the keynote speakers for RV 2025: Thomas Henzinger (ISTA), Nils Jansen (Ruhr-University Bochum), Ankush Desai (Amazon Web Services), and Daniela Micucci (University of Milano-Bicocca).

We're Hosting RV 2025 at TU Graz
Mar 04, 2025

We're Hosting RV 2025 at TU Graz

This September we are thrilled to host the 25th International Conference on Runtime Verification (RV) right here at TU Graz (Sep 15-19, Inffeldgasse 11).

TU Graz Article: Autonomous AI Assistant to Build Nanostructures
Jan 16, 2025

TU Graz Article: Autonomous AI Assistant to Build Nanostructures

TU Graz published a feature on our group’s work building autonomous AI systems for molecular nanostructure assembly. A joint project with physics, chemistry, and mathematics, funded by the Austrian Science Fund.