Overview: According to the EASA website, the European Union Aviation Safety Agency (EASA) has launched a consultation period for its first regulatory proposal concerning artificial intelligence (AI) in aviation. This marks the first step in the regulatory task RMT 0742. The second Notice of Proposed Rulemaking (NPA) is scheduled to be released within 2026, incorporating this general certification framework into the regulations of relevant aviation sectors. From conceptual papers to specific clauses, the regulation of civil aviation AI in Europe is moving towards implementation.
According to a news release on the EASA website, this proposal, which is now open for public comment, represents the initial outcome of RMT 0742 and aims to help the aviation industry prepare for future AI requirements. It focuses on two levels of AI: Level 1 AI, which is AI-based assistance, and Level 2 AI, which is human-AI teaming. The proposal covers guidance on AI assurance, human factors, and ethical considerations, and applies to data-driven AI systems, including supervised and unsupervised machine learning.
EASA also stated that this framework will be further expanded in the future to include technologies such as reinforcement learning and knowledge-based approaches, leaving room for more advanced AI applications.
According to an industry column, EASA plans to release its second sector-specific NPA in 2026, incorporating AI certification requirements into existing certification standards for areas such as operations, air traffic management, and maintenance, thereby providing more detailed operational guidance than the current framework.
Complementing this is the industry's outline of EASA's "four key reliability attributes" – attributes that every AI system seeking certification must possess. The first of these is explainability: the operator must be able to understand why the AI made a particular decision. This set of attributes is becoming the "common language" in discussions about AI certification in Europe.
According to an official announcement by EASA, the proposed Issue 03 of its AI Concept Paper was recently released. This is considered a significant step in implementing AI Roadmap 2.0, expanding guidance on reinforcement learning, symbolic AI, and Level 3 AI applications focused on advanced automation. Stakeholders can provide feedback during a 10-week consultation period, with a deadline of August 12, 2026.
At the same time, EASA announced that it will host the "AI Olympics" during AI Days 2026, which will be held in Cologne in September. Organizations developing AI systems for the aviation industry can submit their proposals to compete for a place among the top 10 cases selected for on-site demonstrations. The deadline for submissions is July 5, and only the first 50 registrations will be considered.
The conceptual paper sets the direction, the first NPA establishes the framework, and the second NPA delves into specific areas – EASA is shaping 2026 into a “year of finalizing regulations.” For manufacturers and operators, the consultation period before August 12th represents one of the few opportunities to directly influence the wording of the regulations. For observers, the value of EASA’s approach lies in breaking down the abstract concept of “trustworthy AI” into a list of verifiable attributes. The more specific the regulations, the more predictable the path for demonstrating compliance for AI systems, and the lower the communication costs for both those within and outside the industry. How the second NPA will be incorporated into maintenance and air traffic management regulations is worth monitoring.