PanelMODELS 2026

About

Traditionally, MODELS hosts a plenary panel on a trending topic, bringing together experts for an in-depth discussion of a particular topic.

Summary

In this edition, panelists from both academia and industry will address and discuss the following topic: “If AI eats software, how can we ensure the result is safe & secure?”.

The advantages of model-driven engineering practices have been proven in several application domains, where the software systems developed must satisfy stringent safety and security requirements. Modeling languages, code and test generation, or formal verification approaches have been designed to aid engineers in creating complex systems with strong guarantees.

However, the capabilities of foundational models and agentic development approaches have reached a point where these tools can autonomously perform even advanced software engineering tasks. They will probably not make software and systems engineers obsolete, as some claim, but organizations should rethink their engineering practices. Moreover, researchers should question some of the assumptions that guided our field. Are modeling languages designed for humans suitable for AI agents? Are AI agents producing the same kind of design faults as engineers? Can current assurance methods keep up with the speed and scale of crowds of autonomously operating agents?

On the one hand, the capabilities to process vast amounts of unstructured artifacts and to access the distilled information produced over the last decades open up numerous new approaches in systems and software engineering. On the other hand, the inherent statistical nature and lack of traceability could pose trust issues with AI-assisted or agentic development approaches. The formal, certifiable assurance offered by model- driven engineering methods could complement AI-assisted methods. The panel will discuss good practices for combining the best of both worlds to build trustworthy engineering systems.