What this session covered
This session examined what a genuinely human-centred approach to AI meant for both training and live air traffic control, rather than treating AI as a technology drop-in. Chaired by Tim Arel — Chair of CANSO and Chief Operating Officer of the FAA Air Traffic Organization — it brought Anthony Mackay, VP and Chief Safety & Quality Officer at NAV CANADA, together with Greg Pile and Terry Biggio to draw the line between a system that is genuinely human-centred and one that merely has a human in the loop. The panel set out why that distinction matters for operational safety, how AI tools were changing what controllers needed to learn and how they were trained, and how ANSPs were building safety and quality assurance around tools that behave differently from traditional deterministic systems. The operational question it answered was practical: under what conditions does AI in the control room enhance controller effectiveness rather than erode it?
Why it mattered
By 2026, AI had moved from concept to presence in ATC environments — in decision-support tools and in the simulation systems used to train controllers. The framing the panel insisted on was that AI in ATC is a human-performance question before it is a technology question: how tools are designed, introduced, and governed determines whether they raise or undermine controller effectiveness and, with it, safety. That mattered against a backdrop of the largest FAA controller hiring programme in a generation and persistent ATCO staffing constraints across European ANSPs, where any tool that changes training requirements or controller workload carries directly into capacity and safety. A NAV CANADA safety leader chairing alongside the CANSO Chair signalled that the industry was treating human-centred AI as a safety-assurance discipline, not a procurement preference.
Key takeaways for ATM operators
- "Human-centred" is not the same as "human in the loop." The panel drew a hard distinction: a system designed around the controller's cognition and authority is different from one that simply leaves a person nominally in the loop, and the difference is decisive for operational safety.
- AI changes training, not just tools. As AI decision-support entered operations, it changed what controllers needed to learn and what simulator environments had to replicate to stay relevant — a training-pipeline implication, not only a technology upgrade.
- Assurance must fit non-deterministic systems. ANSPs were building safety and quality frameworks around tools that do not behave like traditional deterministic systems, drawing on NAV CANADA's operational and safety-leadership experience.
Frequently asked questions
Who spoke on the human-centred AI in ATC panel at Airspace World 2026?
The session was chaired by Tim Arel, Chair of CANSO and Chief Operating Officer of the FAA Air Traffic Organization. The panel featured Anthony Mackay, VP and Chief Safety & Quality Officer at NAV CANADA, alongside Greg Pile and Terry Biggio.
When and where did this session take place at ASW 2026?
The session took place on Wednesday 27 May 2026 from 12:00 to 12:50 local Lisbon time (WEST, UTC+1) in the Boeing Theatre at FIL — Feira Internacional de Lisboa, the Parque das Nações venue that hosted Airspace World 2026 from 26 to 28 May 2026.
What does "human-centred AI" mean in air traffic control?
Human-centred AI describes a system designed around the controller's cognition, workload, and authority, rather than one that merely keeps a person nominally in the loop. The panel argued the distinction is decisive for operational safety, because a tool that is genuinely human-centred supports the controller's decision-making instead of displacing or degrading it.
How does AI change controller training?
As AI decision-support tools entered live operations, they changed what controllers needed to learn and how they were trained, including what simulator environments had to replicate to remain relevant. The panel treated this as a training-pipeline question — a shift in competencies and validation — rather than a simple technology upgrade.
How do ANSPs assure the safety of AI tools that are not deterministic?
ANSPs were building safety and quality assurance frameworks specifically for tools that behave differently from traditional deterministic systems, where the same input does not always produce an identical output. Anthony Mackay brought NAV CANADA's safety-leadership experience to how a major ANSP integrates AI while maintaining its safety obligations, which was central to the session's discussion of assurance.