AI Safety

The department of AI Safety conducts research on how artificial intelligence behave in real-world settings, and how these systems can be tested, evaluated and made more reliable.

The department is part of the Centre for AI Security and Safety, a collaborative research centre between Simula Research Laboratory and SimulaMet.

AI is increasingly used in healthcare, public services, education, and critical infrastructure. While these technologies can create significant benefits, they can also fail in unpredictable ways once they interact with real users, changing environments, and different regulatory and societal contexts.

Much of today’s AI safety research focuses on how models are trained. Researchers in the AI Safety department study what happens after deployment: how systems behave in practice, how they can be tested for vulnerabilities and unintended harmful behaviour, and how organisations can identify and manage risks.

"Most safety work happens where models are built. We work where they are used."

Michael Riegler, Head of Department

Projects

The department of AI Safety was established in spring 2026. More information about the research and active projects is coming soon.