WASHINGTON - The Federal Aviation Administration (FAA) has released a document it says will help assure safety in the use of artificial intelligence (AI) in airborne and ground systems. The roadmap was developed following two years of meetings with industry stakeholders.
In 2022 and 2023, the FAA held a series of meetings to gauge industry priorities and hear opinions.
The agency notes that the aircraft and avionics industry has its eyes on AI image processing tasks, among others, but because artificial intelligence falls outside normal operations - i.e. a designer who can readily explain every aspect of system design - it must take a new approach.
The FAA established seven guiding principles to provide safety assurance in AI systems in aircraft and ground operations, including the need to use existing civil aviation processes and methods. The agency noted the focus of AI technology needs to be on safety assurances and what safety enhancements can be made. In addition, the FAA cautions against anthropomorphizing the technology.
"The system designer must delineate the responsibilities that are assigned to human beings as compared to the requirements that are assigned to systems and tools and must do so in a manner consistent with applicable aviation regulatory requirements and international standards," the FAA writes in its roadmap. "The responsibility for systems to meet their requirements rests with the system designer and AI developer, not the AI itself."
The FAA says it is necessary to differentiate between learned AI and learning AI. The FAA makes the distinction by categorizing learned AI systems as static in their operating environment while learning models continually incorporate adaptations into their reasoning.
"Learning AI implementations may adapt in a manner that degrades performance, ultimately weakening their original safety profile," the FAA says in its document. "Cases in which a system learns anomalous, ungeneralizable, or inaccurate information will require new mitigation strategies. Learning systems may necessitate new regulations to assure the continued safety of the evolving system, as for active monitoring of performance or recurrent certification."
Other principles include the need to take an incremental approach to AI adoption and to leverage the existing safety continuum.
"The highest level of safety is expected for scheduled passenger service, transitioning to a lower safety threshold for research experimental flights and drone operations. We can gain experience with AI in experimental aircraft, without trying to provide the assurance that would be required for that AI to be used in scheduled passenger air carrier operations," the agency writes. "Small uncrewed aircraft also provide ideal vehicles and operations to gain early experience which can be used to further inform future versions of this roadmap. The experience that is gained can inform safety assurance methods relevant to other applications and safety objectives."
Finally, the FAA's guidelines include the need to leverage industry consensus standards.
"Industry standards can promote global regulatory harmonization if widely adopted and can be agile in adapting to the changing technology," writes the FAA. "The aviation community should consider the standards developed for other applications outside of aviation and develop aviation-specific standards."
The agency notes that its roadmap is very much a “living document” and will be augmented as both capabilities and industry grow.
“This document reflects a point-in-time snapshot of a fast-paced and evolving technology,” says the FAA. “Changes to this roadmap will be implemented based on experience, standards development, and research. It is important to recognize that the maturation of the technology is occurring in a much broader community than aviation and following the further developments outside of aviation is as critical as advancing the activities outlined in this roadmap within the aviation community.”
The FAA’s Roadmap for AI Safety Assurance document is available here: https://www.faa.gov/aircraft/air_cert/step/roadmap_for_AI_safety_assurance
