Can Scientific AI Be Certified? Professor Drikakis Outlines Flightline Pathway.

The future of air transport requires new tools. The Bauhaus Luftfahrt Symposium 2025 convened global leaders to discuss this path. The event, held on 18–19 November at Munich Airport, marked the think tank’s 20th anniversary. Its theme was "From 2005 to 2070 – Our Journey to the Future of Aviation."

Operationalising Trustworthy AI

Professor Dimitris Drikakis, Dean of Sciences and Engineering at the University of Nicosia, delivered a key invited talk. His presentation focused on a critical industry challenge: turning rapid AI predictions into flightline-ready evidence. The talk, titled “From Prototype to Flightline: Operationalising AI Safely and at Scale,” centered on frameworks for safe, trustworthy, and certifiable AI.

Professor Drikakis addressed the "Transformative Role of AI: Potential and Pathways" theme. He argued that transforming aviation hinges on Scientific AI aviation. This approach uses models that are inherently testable and verifiable.

The Verification-Driven Pathway

His core message outlined a verification-driven pathway for advanced computational models. This pathway bridges the gap between research prototypes and operational deployment. It ensures that AI acceleration is paired with robust safety assurance.

Key components of this framework include:

• Physics-informed models: These embed conservation laws and constraints. This reduces unreliable extrapolations in safety-critical regimes.
• Transformer surrogate models: These are designed for rapid scenario exploration. They preserve the traceability of assumptions and data provenance.
• Digital twins flightline: These support continuous alignment. They link models, test evidence, and operational realities.

AI Assurance and Regulatory Realism

The aviation industry relies on safety-critical systems. Therefore, AI-based systems must meet the same strict certification standards as traditional software. Professor Drikakis emphasized the need for rigorous Verification and Validation (V&V). This includes disciplined data scrutiny and uncertainty-aware Key Performance Indicators (KPIs).

Regulatory bodies like the EASA and the FAA are actively developing new frameworks. The traditional V-model process is not fully suited for iterative AI development. EASA, for instance, introduced the W-shaped process to run parallel to the V-model. This adds dedicated requirements for data management and model training.

Guidance for advanced automation (Level 3 AI) is expected from EASA in 2025. This highlights the urgency of establishing certifiable pathways. The Bauhaus Luftfahrt Symposium 2025 program reflected this focus on AI certification pathway and assurance.

Impact on Climate-Neutral Aviation

The Symposium’s broader focus was the journey to climate-neutral aviation. Speakers from Airbus, Lufthansa Group, and MIT contributed to a program linking sustainability, economics, and frontier engineering.

Scientific AI is a system-level enabler for this transition. It can optimize propulsion innovation and alternative fuels research. By providing fast, testable predictions, physics-informed models accelerate the design of more efficient aircraft components.

• Opportunity: AI can accelerate the exploration of new, sustainable aircraft designs.
• Challenge: Speed must be balanced with auditable, decision-grade performance claims.

This framing treats AI assurance as a first-order constraint, not an afterthought. It ensures that innovation contributes meaningfully to both safety and sustainability goals in commercial aviation news. The goal is to make AI operational and trustworthy at scale.