Federal Aviation Administration officials have denied that the agency's SMART artificial intelligence system played any role in a nationwide ground stop that disrupted air travel starting Friday. The denial comes as questions mount about the system's safety oversight and its place in the agency's air traffic operations.
SMART, which stands for Strategic Management of Airspace, Routes and Trajectories, combines about 200 data streams into a single platform. The system is designed to predict congestion and to handle weather more precisely, and it recommends actions to air traffic personnel. Officials emphasize that the software does the computational work while humans make the final flight decisions. The startup Air Space Intelligence, which developed the product, has also said its technology did not cause the ground stop.
The system has drawn scrutiny since its rollout. It cost $875 million and launched nationwide on Nov. 10, making it one of the most ambitious automation efforts in the agency's recent history. Last month, House Democrats asked the Government Accountability Office to audit SMART. Their concerns center on safety, the strength of the system's guardrails, and the degree to which air traffic controllers have been consulted in its design and use. Their request reflects a broader debate over how far automated tools should go in a system where errors can carry serious consequences.
Supporters point to early results. In a test around the Washington area, the system rerouted 180 flights, saving roughly 10 minutes and 80 gallons of fuel per flight. In one recommendation, a Tampa-to-Dallas flight was shortened by 42 minutes. Those gains form the core argument for the technology: fewer delays, lower fuel use, and more efficient routing through crowded airspace. The developers frame the tool as a planning aid rather than an autopilot for the national airspace.
Critics, including former FAA officials, have urged that the agency verify the system's outputs carefully to guard against AI errors. Concern has focused on the risk of so-called hallucinations, in which an AI system produces confident but incorrect information. Such a failure would be especially worrying in aviation, where recommendations must be reliable under time pressure and where a single faulty suggestion could ripple across many flights.
The debate also unfolds against a persistent staffing problem. A union representing controllers reports roughly 14,000 controllers on the job, and the shortage continues to strain the system that SMART is meant to support. Air traffic management is already stretched, so any tool that shifts workload or alters decision-making draws close attention from the people who must act on its outputs.
The investigation into the ground stop's causes has not been completed, and officials have not provided a full account of how long the disruption lasted. Air Space Intelligence's denial and the FAA's statement leave the question of responsibility unresolved for now. Internationally, the picture is different: no other country has a working equivalent of SMART, and Eurocontrol, the European air traffic organization, is still testing similar approaches. The American effort therefore remains a pioneering and closely watched experiment.