Overview: 2026 is considered by the industry to be a pivotal year for the transition of civil aviation artificial intelligence (AI) regulation from "concept" to "formal regulations." According to an industry column, regulatory bodies on both sides of the Atlantic are building their respective AI standard frameworks. Among them, the European Union Aviation Safety Agency (EASA) has a relatively clear and public roadmap: the first regulatory proposal is open for public comment, and the second Non-Binding Technical Opinion (NPA) is planned to be released within the year. This article, based on publicly available information, provides an overview of this roadmap.
Looking back at EASA’s approach, we can see a three-stage path. The first stage involves the AI Roadmap 2.0 and the AI concept paper, which serve as guiding documents to define concepts and expectations. The second stage is the first regulatory proposal – the initial step in the RMT 0742 rule-making task. According to EASA’s official website, this proposal covers AI assurance, human factors, and ethical guidance, and focuses on data-driven AI systems (supervised and unsupervised machine learning) to help the industry prepare for the future requirements of Level 1 AI (AI-assisted) and Level 2 AI (human-AI teaming). The third stage is the second NPA (Notice of Proposed Amendments), which will deploy the general framework into the regulations of relevant aviation fields by 2026.
According to materials from the EASA AI Days technical conference, EASA plans to implement AI and machine learning in stages, categorized by "levels." The guidance plans for Level 1 and Level 2 AI and machine learning are scheduled to be completed by 2026, while the guidance for Level 3 is planned for completion around 2028. This timeline indicates that applications with lower risk and greater human oversight will be the first to receive regulations, while more autonomous Level 3 applications will wait for more comprehensive technological and regulatory preparation.
Complementing the timeline are the “four key attributes of trustworthiness” identified by industry experts. According to an article, EASA requires that each AI system seeking certification possess these attributes, the first of which is explainability – meaning that the operator can understand why the AI made a particular decision.
For the Maintenance, Repair, and Overhaul (MRO) sector, the most direct impact comes from the second NPA. According to an industry column, EASA plans to use this as an opportunity to incorporate AI certification requirements into existing certification standards for areas such as maintenance. This means that AI-assisted inspection and predictive maintenance tools, if they are to be used in the European market, may need to meet the same reliability requirements as those in the airworthiness sector. Combined with the operations and air traffic management sectors, the scope of the regulations will extend from "manufacturing" to "usage.
Considering all the above information, 2026 is significant because it marks the first time that the technical classification (Level 1/2/3) and the regulatory timeline (the two NPAs) are aligned within the same year. It should be noted that the specific timeline for FAA-related documents is limited in the publicly available information used in this article, and any related progress should be based on official announcements. It can be expected that cross-regional coordination will be a key focus in the future. If the same AI system is to be certified in multiple markets simultaneously, the degree of convergence in regulations will directly affect compliance costs. Furthermore, how the second NPA addresses maintenance and air traffic control scenarios will be the first test of the practical application of this framework.