Predictive maintenance, flight data, safety performance analysis, and an AI-based flight safety management system. These are just a few examples of how artificial intelligence can be utilized in aviation. However, it is not unimportant how an operator uses this technology, which is why the European Aviation Safety Agency (EASA) recently published a proposal outlining the conditions for the application of artificial intelligence in aviation.
The subtitle of the EASA proposal, which refers to the trustworthiness of artificial intelligence, is telling, meaning that the agency is also aware of the problems and doubts surrounding artificial intelligence, such as the lack of data purity, which can also endanger flight safety. In addition to flight safety considerations, the publication of the proposal is also unavoidable due to the European Union’s so-called AI Act and the legislative changes it has prompted.
Building on these requirements, operators using artificial intelligence will also have to describe a concept of operations for their AI-based systems in the future. This concept will define which members of the crew will use it (e.g., remote pilot), who is considered an end user, what tasks the AI performs, how the tasks are divided between the AI and the end user, how the end user and artificial intelligence work together, and what the limitations and conditions for the use of the applied AI are.
Another important element of the proposal is the appropriate classification of artificial intelligence systems according to the extent to which the end user can override AI decisions, and a risk analysis must be carried out in accordance with the classification.
In the not-so-distant future, the description of the applied AI will also be part of the operating manuals, and personnel training may be supplemented with elements related to artificial intelligence.
The proposal is available here.