INSIGHT
As global air traffic volumes continue to climb through 2026, the demands on Air Traffic Management (ATM) systems have grown significantly. Traditional systems are being pushed to their limits and need more innovative solutions to maintain safety, efficiency and sustainability.
Artificial Intelligence (AI) and Machine Learning (ML) are at the forefront of this transformation, and the regulatory framework around them has matured rapidly. EASA’s AI Concept Paper has progressed through successive issues, EUROCONTROL has published guidance on the safe and trustworthy use of AI in ATM, and EUROCAE WG-114 and SAE G-34 have continued to develop the assurance standards that will allow these technologies to be certified. This article looks at how those threads are now coming together.
ATM services today rely on a blend of surveillance, communications, navigation and — above all — human expertise to guide aircraft safely through controlled airspace. Surveillance itself spans primary and secondary radar, Mode S, multilateration (WAM) and space- and terrestrial-based ADS-B; the underlying technologies vary by region and operating environment. The pressure point is less the equipment and more the service: as traffic volumes grow and new entrants — uncrewed aircraft, higher-airspace operations — arrive, the demands on controllers, supervisors, flow managers and the wider operations room grow with them.
The near-term opportunity for AI and ML, then, sits largely on the human side of the service rather than inside certified safety-critical systems. Decision-support tools, workload prediction, briefing and shift-handover support, conformance monitoring, training, post-operations analytics and back-office functions all stand to benefit well before AI is woven deeply into the on-line ATM platform itself.
AI algorithms can process vast amounts of historical and real-time data to predict traffic patterns, weather conditions and potential disruption. This foresight enables systems to anticipate issues before they arise, allowing for smoother operations and more efficient use of airspace.
AI-driven tools can assist controllers by providing real-time insight and recommendations. These tools reduce workload and support faster, better-informed decisions, while keeping the human firmly in command of the safety-critical loop.
Machine learning excels at recognising patterns — and, crucially, deviations from them. Applied to surveillance and system health data, ML can flag emerging anomalies earlier than conventional monitoring, supporting a more proactive safety posture.
The safe adoption of AI in a safety-of-life domain depends on trustworthy, explainable and verifiable systems. The framework that will enable this is now visible on several fronts:
Because most early AI use will sit on the human side of the service, organisations need to establish a clear company AI policy covering how staff — operational and support — are permitted to use AI and machine-learning tools. The policy needs to be specific about which tools are sanctioned for which tasks, what data may be put into them, and how outputs are recorded.
The single most important principle is that responsibility for the output stays with the human. AI and ML tools can accelerate analysis, drafting, screening and pattern-spotting; they cannot take accountability. Safety Management staff, in particular, must take the time to review AI outputs properly, understand the basis on which they were produced, and stand behind the conclusion they sign off — whether that is in a safety assessment, an investigation, a change decision or a board paper. The same discipline applies in engineering, operations, training and back-office functions: AI is a tool used by a qualified person, not a substitute for one.
A workable policy typically covers permitted use cases, prohibited use cases, data handling and confidentiality, review and sign-off expectations, record-keeping of AI-assisted work, and the interaction with Part-IS information-security controls.
For ATM/ANS providers operating under the European regulatory framework, the practical question has shifted from “can we use AI?” to “how do we evidence that we are using it responsibly?” That means treating AI-based functions as classified changes under EU 2017/373 change procedures, applying the design-assurance and conformity-assessment regime of Part-DPO (EU 2023/1768 & 1769) where AI is embedded in ATM/ANS equipment, integrating assurance into the existing safety case rather than alongside it, exercising the information-security controls now required under Part-IS, and underpinning the lot with a company AI policy that keeps the human accountable.
For providers not operating under the European regulations, the picture is broadly the same — just without the regulatory anchors. The same disciplines apply: classify and control AI-based functions as managed changes under the local SMS, document the assurance argument, secure the data and models, and put a company AI policy in place that defines permitted use and keeps a qualified human accountable for every output. The work of EUROCONTROL, EASA, EUROCAE WG-114 and SAE G-34 remains a valuable reference point internationally, with parallel activity from ICAO and the FAA helping to align practice across regions.
AVISU’s standardisation experience, combined with our regulatory and safety practice, positions us to help customers translate the emerging framework into deployable, defensible products — whether they sit inside the EASA system or outside it.
EASA’s new Design and Production Organisation regime under (EU) 2023/1768 and 2023/1769 is here. From the Cologne workshops to the October 2028 transition deadline — what industry needs to watch.
Air traffic and aviation analytics are transforming how flow is managed — improving both efficiency and safety across the industry.
Safety Management Systems provide a structured, proactive approach to managing risk — the backbone of a safer airspace for all.
For more information on our services, speak to one of our experts on 01463 554024 or email contact@avisu.co.uk.