Lawmakers Seek Pause of FAA’s $875 Million Washington AI Tool

An artificial-intelligence tool is already supporting flight-management decisions around Washington, but more than two dozen lawmakers say the FAA has not shared its safety evaluation, testing reports or personnel-training details with them. They are asking the agency to pause the system and provide answers by Oct. 19.

The dispute involves an $875 million, 12-year contract and a safety-critical operating environment, yet it does not establish that the FAA conducted no testing or training. The immediate issue is whether the agency has adequately demonstrated and communicated the system’s validation, operating limits and human-oversight structure before expanding its use. Representative Alexandria Ocasio-Cortez summarized the concern bluntly: “Air traffic control leaves zero margin for error.”

The FAA began using the software in the Washington area in September. The system is formally called Strategic Management of Airspace, Routes, and Trajectories, or SMART. In their request for a pause, lawmakers asked what evaluations preceded deployment, how aviation personnel were prepared to use its recommendations and what standards would govern any wider rollout.

SMART is not an autonomous air traffic controller. It does not issue control instructions to aircraft or take responsibility for keeping them separated in flight. Transportation Secretary Sean Duffy said controllers will remain in control of the airspace and described the software as an additional decision-support tool.

What the software actually does

SMART analyzes airline schedules, weather, airport capacity, airspace conditions and other operational constraints. The Transportation Department says it draws from more than 200 data streams, builds visualizations and recommends scheduling options to operations experts. Its predictive models are intended to identify developing traffic conflicts and congestion before those conditions fully materialize.

That distinction matters. The software operates at the planning and traffic-flow level, where specialists coordinate how the overall airspace system absorbs weather, closures and demand. Human operators decide whether to use its recommendations. The initial operational focus is rescheduling flights canceled by bad weather, rather than controlling individual aircraft in the sky.

Potential later applications include coordinating schedules and trajectories before departure. FAA Administrator Bryan Bedford has argued that better decision-making tools should lead to better decisions, while the agency plans to expand SMART gradually beyond the Washington region.

The contract awarded to Air Space Intelligence in June covers SMART and development of a Flow Management Data and Services system intended to replace the existing platform at the FAA’s Air Traffic Control System Command Center in Virginia. That broader scope helps explain the $875 million ceiling: the agreement is a 12-year software and system-modernization commitment, not simply the purchase price of the limited Washington deployment.

The questions behind the requested pause

For predictive software in aviation, keeping a human in charge is essential but not by itself a complete safety case. Operators also need to understand what a recommendation represents, when its data may be incomplete or stale, how the system behaves during unusual conditions and when established procedures should take precedence. Training and interface design matter because a tool can influence decisions without directly controlling an aircraft.

The lawmakers therefore want the FAA to identify the evaluations completed before deployment, explain system performance during severe weather or outages and disclose whether any safety-review requirements were waived during contracting. They also asked what input came from airport authorities, airlines and aviation safety workers; what training personnel received; and how SMART recommendations are supposed to affect human decisions.

A third group of questions concerns measurement and accountability. The FAA was asked to state how it will evaluate performance during the initial operating period, what standards SMART must meet before broader expansion and who is responsible if one of its recommendations contributes to a safety incident.

Those questions define the practical boundary between deploying a promising planning aid and proving that it belongs in routine aviation operations. They also address the central concern for flyers uneasy about AI entering air traffic management: predictive analytics may shape consequential choices even when the machine never has direct authority over an airplane.

The Oct. 19 deadline now gives the FAA a near-term opportunity to clarify what validation and training occurred, what evidence it can make public and which performance gates will precede expansion. No agency response to those questions is established here. Until one arrives, the unresolved issue is not whether SMART autonomously flies or directs aircraft it does not but whether the safety case for its advisory role has been demonstrated with enough transparency for a nationwide public system.

By Thomas Caldwell — AMI’s senior editor for mechanical and mobility engineering, covering vehicle electronics, systems integration, electrification, chassis systems, propulsion, and safety policy.