Dr Zhibao Mian Presents Generative AI Safety Research at Computing Conference 2026
- Aug 2
- 1 min read
The Dependable Intelligent Systems (DEIS) Research Centre was delighted to be represented at Computing Conference 2026 by Dr Zhibao Mian.
Dr Mian presented “Leveraging Generative Models to Produce Safety Artifacts” during Session 6: Applied AI on 9 July 2026. The paper was co-authored with Dr Koorosh Aslansefat and investigates how generative AI can support the production of safety artefacts from visual scenes.
Safety engineers often need to transform complex operational scenarios into structured analyses that identify hazards, causes and potential controls. This process requires specialist knowledge and can demand considerable manual effort. The DEIS research asks how generative models might assist with the early stages of that work while retaining the structure and transparency required for engineering review.
A key example examines an autonomous-driving scene. The approach interprets the objects and relationships visible in the scene, connects them with safety knowledge and generates a fault-tree representation. This provides an explicit artefact that engineers can examine rather than relying solely on a natural-language answer.
The research forms part of an emerging DEIS theme led by Dr Mian: developing AI systems that can reason in ways closer to safety experts and contribute useful, reviewable evidence to safety analysis. The long-term aim is to reduce manual effort without weakening the discipline, traceability or human oversight on which dependable engineering relies.
Presenting the work at Computing Conference 2026 provided an opportunity to share the approach with the wider applied-AI community and discuss both its promise and its present limitations. These conversations are important as researchers establish where generative AI can genuinely support safety-critical development and where stronger reasoning, validation and assurance methods are still required.
Original LinkedIn update:
Comments