AI in Aviation: The Future is Now
The concept of using artificial intelligence (AI) in aviation is extremely appealing. In a technologically-perfect world, one can imagine AI handling everything from robotic aircraft maintenance to self-flying planes.
Such an idea seems like the stuff of science fiction. But make no mistake: AI is already making its way into flight. For example, Sikorsky’s fully autonomous uncrewed S-70UAS U-Hawk cargo helicopter is currently under development. Designed to be flown by onboard computers using the company’s MATRIX flight autonomy system, the U-Hawk has no cockpit whatsoever. Instead, that space has been added to the U-Hawk’s cargo space, which is accessed by front-opening clamshell doors and a cargo ramp where the cockpit traditionally sits.
This is just one example of AI in aviation, which seems likely to deliver substantial benefits in safety, efficiency, and enhanced system management. The question is, where do we stand now, and what challenges lie ahead? This article will explore that and other questions, with informed input from experts in the aviation industry.
Regulators Diving In
With AI already making its way into aviation, regulators are doing their best to keep up. In America, “the FAA released some brief guidance on AI in aviation maintenance in its ‘Roadmap for Artificial Intelligence Safety Assurance’,” said Rob Mather, VP of Aerospace & Defense with IFS, an ERP platform provider whose markets include aviation. “But overall, the technology is advancing far more quickly than the regulations surrounding it, with a lack of clarity from regulators and the industry as a whole.”
In Europe, EASA’s first regulatory proposal on ‘Artificial Intelligence for Aviation’ was released on November 10, 2025. The goal of the process is “to provide the industry with technical guidance on how to set the ‘AI trustworthiness’ in line with requirements for high-risk AI systems that are contained in the EU AI Act (Regulation (EU) 2024/1689),” said easa.europa.eu. “The NPA is now open for public consultation for 3 months and your comments at this stage are very important.”
In Europe, the European Organisation for Civil Aviation Equipment (EUROCAE) is the EU’s organization that deals with aviation standardisation, for airborne and ground systems/equipment.
Mark Roboff is Co-Founder of SkyThread, a data-sharing network built specifically for the commercial aviation industry. He was also the founding chairman of SAE G-34, which is in a partnership with EUROCAE WG-114. It is the joint G34/WG114 aerospace standards committee that is working with global industry and regulators to devise a means of compliance for the certification of machine learning (ML, a subset of AI) into aircraft and air traffic management systems. “In alignment with major industry and regulatory leaders, the standards committee is on track to publish its first recommended guidance, ARP-6983, which will detail assurance methods for building and integrating trustworthy AI into aerospace systems up to a Design Assurance Level (DAL) C safety standard,” Roboff told Aerospace Innovations magazine. “Our work is aligned with the EASA AI Roadmap, associated published concept guidance, and other supporting material, and is designed to also support FAA guidance on the subject.”
Thuc Nguyen is a Technical Programme Manager at EUROCAE. He oversees several working groups in the domains of Information Technology, Software, and Artificial Intelligence.
“WG-114 has successfully published some valuable deliverables,” said Nguyen. They include:
ER-022 “Artificial Intelligence in Aeronautical Safety-Related Systems Statement of concerns” (October 2021): This document serves as that gap analysis and provides a list of concerns that need to be addressed in order to produce a future means of compliance.
ER-027 “Artificial Intelligence in Aeronautical Safety-Related Systems Taxonomy” (December 2024): This publication establishes a comprehensive taxonomy of Artificial Intelligence in aviation, supporting the effective implementation of the ED-324 guidelines.
Several more standards are under development by EUROCAE. They include:
ER-xxx “Artificial Intelligence in Aeronautical Safety-Related Systems Use Cases Considerations”: This document highlights practical use cases from the industry, demonstrating how the concepts outlined in the main technical standard, ED-324, are being applied in real-world scenarios.
ED-324 “Process Standard for Development and Certification Approval of Aeronautical Products Implementing AI”: This is EUROCAE’s principal AI in aviation technical document. It establishes industry best practices for the development and certification of AI embedded in aerial vehicles and ground equipment. This foundational standard provides a structured framework to ensure that AI technologies integrated into aviation systems meet rigorous safety and reliability requirements, paving the way for their regulatory acceptance and operational deployment.
Clearly, the regulators are doing their best to dive into AI in aviation. Here’s the problem:
“Aviation is a dramatically fragmented and siloed industry, so there is no one set of standards for AI,” Roboff observed. “The standards I have worked on involve how to certify machine learning into avionics and air traffic management systems and therefore are strictly concerned with the design and engineering part of our industry. There are standards for certain segments in the operational space, such as standards for maintenance programs and standards for certain types of data structures, but these do not yet involve AI, nor does AI’s application in the operational space necessarily require the creation and use of standards, unlike the design and engineering space.”
To make matters worse, “there’s a definite gap growing between the development of AI applications in aviation maintenance and the regulations required,” said Mather. Very few, if any, decisions have been made for AI standards in aviation maintenance. While the FAA has begun to recognize the need to introduce AI into aviation operations, its initial roadmap has been cautious about widespread usage. Moving forward, more needs to be done by regulators to create a certified AI model. Standards need to be set to allow AI decision-making to be verified and validated, and aviation maintenance organizations to be able to take advantage of the fast-growing number of AI-enabled applications.”
Some Real Concerns
At a fundamental level, the biggest issue about using AI in aviation is the relatively new and not-fully-understood nature of this transformative technology. As a result, there are real concerns among the experts we spoke with that AI as it currently stands may not be mature enough for widespread deployment in aviation.
“From a safety standpoint, there is growing concern that existing development assurance standards may not be fully suitable for AI/ML-based solutions,” said Nguyen. “These standards typically rely on a predefined set of activities and demonstrations, which certification authorities use as evidence of compliance. However, there is broad consensus that the currently accepted means of compliance for systems, software, and hardware do not adequately address the unique characteristics of certain AI/ML techniques.”
Here’s the problem: “AI and ML methodologies often diverge significantly from the traditional development life cycle assumed by existing assurance frameworks,” he told Aerospace Innovations. “Their data-driven, iterative nature challenges conventional validation and verification processes, making it difficult to apply legacy standards effectively. As a result, there is a pressing need to adapt or develop new assurance approaches that can provide appropriate safety guarantees for AI/ML-enabled systems in aviation.”
“Generally speaking, the key issue with the application of AI in aviation is a matter of trust,” said Roboff. “How can we trust the AI to perform in the same manner that we can trust certified software to perform? Or how can we trust AI to perform in the same manner as we trust a pilot or a mechanic? The work of my committee, along with its industry and regulatory partners, has been and is to answer these questions.”
“As always, the primary concern in aviation is safety,” Mather noted. “We can’t allow a safety-related incident to occur because of AI. However, IFS can speak directly to AI-related issues with on-the-ground maintenance technicians. Aviation maintenance organizations have employees that work in both regulated and non-regulated roles. When maintenance organizations carry out non-regulated activities such as maintenance planning, or the act of assigning technicians to tasks, those are non-regulated roles so there is no reason or regulation saying they cannot use AI to make them more effective. However, issues arise when we’re talking about regulated activities such as the actual execution of a maintenance activity, because the maintenance technician is fulfilling a licensed and certified role, and there is no means of granting such certification to an AI model today. So, while we can support the technician by AI providing them with information and options, we can’t supplant their ultimate decision-making.”
AI in Aviation Today
The concerns cited above do not denigrate AI’s ability to make positive, productive, and even safety-enhancing contributions to aviation. For instance, the ability of AI-enabled systems to process and analyze massive amounts of data make this technology potentially very useful for Flight Data Monitoring and Safety Management Systems, among others. The accurate tracking of parts and scheduled repairs in the MRO sector is another area where AI can shine, because AI-enabled data systems do not suffer from human fatigue or error.
“This is why non-safety-critical applications of AI are already being widely adopted across various areas of aviation, particularly in enhancing customer experience and optimising internal operations within organisations,” said Nguyen. “As we approach the publication of ED-324, we anticipate a broader range of potential applications (from digital operations to air traffic management) emerging in the near future. This upcoming standard marks a significant step toward enabling the safe and effective integration of AI technologies across both operational and regulatory domains in aviation.”
According to Mather, “there are already several AI use cases within aviation — think maintenance scheduling optimization, error detection and reclassification, and troubleshooting support. Once we have a regulated environment for AI to work in, predictive aircraft maintenance can be used instead of traditional maintenance programs, transforming the way we view maintenance processes moving forward.”
“AI is being used today to aid decision making in certain airline operations, such as maintenance planning and execution,” Roboff agreed. “Where it is being used, its use is strictly limited to being a decision support system. It is not automating a regulated or safety critical process, nor is it making decisions on behalf of a regulated signatory. It is also being used extensively in the commercial realm of airlines in regard to sales and marketing functions, i.e. revenue management, pricing, and loyalty.”
Having said that, Mark Roboff made clear that “AI is not used in any capacity today on board a certified aircraft system. It is not used to automate any element of flight, nor is it used to provide a higher degree of autonomous function that existing automation can provide. However, there are proven examples of where an AI (machine learning) produced algorithm, if integrated onto an airplane, can provide superior performance to a traditional hand-coded algorithm without impacting automation or safety boundaries. Examples include flight path planning and fuel consumption optimization. As a result, we can expect the first use-cases of ‘onboard AI’ to be in these domains. As the technology and its use in industry matures, and as trust in the systems designed to assure its safety is bolstered, we can expect AI to take on more roles involving automation and autonomy.”
The Future of AI in Aviation
Will we ever see AI take the yoke for civilian aircraft? Douglas King, CEO of Epic Aircraft, believes the answer to this question is yes. “Small passenger flying pilotless aircraft with a few passengers is likely to occur in the next 20 years,” he said. “It will take much longer for large aircraft to be flown autonomously.”
Mark Roboff endorses King’s opinion. “I do believe we will see the arrival of self-flying aircraft, and I do believe AI aircraft can be proven statistically and holistically to be safer than the piloted aircraft we rely upon today,” he said. “However, I do not believe we will see self-flying aircraft used ubiquitously for quite some time, and I do believe that starting applications will be in the cargo space.”
“From a standard aviation safety perspective, it is far more likely that self-flying aircraft will first be deployed in cargo operations rather than passenger transport,” said Nguyen. “The primary reason is risk tolerance: Regulators and the public have significantly lower tolerance for risk when human lives are involved. As well, safety requirements for passenger aircraft are exponentially more stringent. Every component of an autonomous flight system, including sensors, software, hardware, integration, and control mechanisms, must demonstrate an exceptionally high level of reliability. Achieving such reliability across all flight conditions, including unexpected weather, system faults, or air traffic conflicts, presents a major technical challenge for an AI-controlled aircraft. Another critical factor is public confidence. Even if autonomous systems can operate safely from a technical standpoint, gaining trust from passengers is a separate hurdle. People will need to see years of safe and proven operations before fully embracing this new mode of transport as they do conventional aircraft.”
Yet science fiction is poised to become science fact sooner than most people would expect. A case in point: “At IFS, we have several industry-leading Advanced Air Mobility (AAM) customers who are right at the forefront of these developments and are reacting to self-flying aircraft very differently. If we look at Joby Aviation, they already have plans to develop pilotless eVTOLS, but initially, they have opted to launch a piloted air taxi service,” Mather noted. “Contrast that with a company like Wisk Aero that is going pilotless from the start and is currently developing its sixth-generation pilotless eVTOL aircraft. So it’s just a matter of when, not if, we will see these self-flying aircraft in operation. It’s important to distinguish between remotely-piloted and self-flying as well and AI has a part to play in both scenarios. We already have remotely-piloted drones in operation delivering cargo — it’s just a matter of scale. As such, we believe we will see self-flying planes for cargo much earlier than for passengers and for passengers, to put a time scale on this, we’re still talking about decades, not imminently.”
Regulation Will Temper Progress
With Joby, Sikorsky, and Wisk Aero pushing ahead with AI-enabled pilotless flight, the advent of AI’s full integration into aviation is now a matter of when, not if. According to the experts we spoke with, the timeline in which this happens will be determined —- or at least constrained — by the regulators, much like a jockey trying to keep a thoroughbred racehorse in check.
“AI in aviation will be subject to strict constraints to ensure safety, security, reliability, and accountability,” said Nguyen.
“These constraints include rigorous requirements for verification and validation, human-in-the-loop oversight, robust system performance, and legal responsibility. The aviation sector is expected to adopt a step-by-step approach, as outlined in the EASA AI Roadmap, to gradually integrate AI technologies. This phased strategy is designed to ensure that each advancement is thoroughly assessed and validated before moving forward, maintaining the sector’s high safety standards while enabling innovation.”
Mather expects the rollout of AI in aviation to occur in three waves. “Wave One is what I would call ‘Human at the Core’, which is happening now,” he said. “These are AI applications we can deploy today where we are supporting humans doing their tasks, but ultimately all the decision-making remains with the human. Wave Two is ‘Human in the Loop’: Decisions can be made by the AI but they are still supervised by a human, who still makes the ultimate decisions. Wave Three, which I call the Certified AI Wave, is where we ultimately start unleashing the true potential of AI, namely allowing AI to do what it is good at independently. To get to Wave Three, we need both technological improvements to AI and a significant shift in regulatory regimes, and these will both take time.”
“It would be senseless to place limits on AI prematurely,” added Roboff. “However, as the industry develops use-cases with AI and continuously demonstrates the safety assurance of AI systems, I expect regulators to move in lockstep to ensure certification and usage processes to be clear and consistent.”
So when will AI in aviation become a full-fledged reality? When it comes to AI-enabled flight technology, the answer is now. What remains to be seen is how long it will take for this technology to reach a truly safe, predictable level that can be certified by regulators with confidence. But again, make no mistake — this day is coming.
By James Careless
