Preparing AI Research for Certification & Airworthiness

Kurzfassung

The certification of AI-based systems in safety-critical domains is evolving rapidly, with frameworks such as the EASA AI Concept Paper, the EASA AI Roadmap and the EU AI Act becoming increasingly concrete. In aviation maintenance, repair and overhaul (MRO), AI is entering a growing range of inspection and maintenance tasks. Yet in research practice, certification is still commonly treated as a downstream step: a model is developed and its performance demonstrated first, and only then does the question of certification arise.

This session opened that sequence up for discussion. Its guiding question is deliberately unresolved: to what extent should research already be prepared for later certification — and where does early preparation help rather than hinder? Some decisions taken at the start of research are certification-relevant and hard to reverse; others can safely be deferred. Preparing too little risks later dead ends; preparing too much can constrain research unnecessarily. Where the line between "prepare now" and "defer safely" should run is not something we aim to settle in advance, but to think through together.

The workshop addresses researchers, developers and certification specialists working on AI for aviation MRO, and lives from the exchange between these perspectives. The outcomes are understood as an evolving basis for discussion, documented and shared with the community