AI Accelerates ATM Software Development but Poses New Assurance Challenges, Egis Warns
Why It MattersAssurance processes, not code generation speed, are likely to become the limiting factor determining how quickly AI-accelerated ATM innovations reach live operations.
Artificial intelligence is reshaping software development in air traffic management, compressing tasks that once took weeks into hours or minutes, but Egis warns the bottleneck is shifting rather than disappearing: faster code generation does not reduce the time needed to verify, validate, safety-assess and gain regulatory acceptance for new capabilities.
In a position piece, Egis argues that verification, validation and safety assessment already account for a substantial share of ATM delivery timelines, and cautions against automating assurance the way AI automates development. If a single AI model generates requirements, code, tests and traceability evidence, the apparent consistency between those artefacts may be misleading — what Egis describes as "AI marking its own homework." Automated verification checks only what it has been instructed to check, leaving integration testing, operational trials and long-duration soak testing essential for exposing unanticipated failure modes.
Egis notes that accountability for safety-significant change will remain with named human decision-makers rather than AI models for the foreseeable future, and warns delivery pressure could lead engineers to defer too readily to machine-generated recommendations. The company points to activity under way at EUROCONTROL, CANSO, EUROCAE and EASA, including EUROCONTROL's FLY AI programme, as evidence the industry is beginning to address responsible AI adoption in safety-critical environments, and states it is contributing through industry forums, working groups and collaborative projects with customers and regulators.
