Who Can Overrule the AI Control Tower?
The FAA awarded Air Space Intelligence an $875 million, 12-year contract and plans to begin testing its SMART airspace-planning software around Washington, DC, on Monday.
Controllers need reliable authority, training, logs, and fallback procedures because a faulty or compromised recommendation could disrupt flights or obscure responsibility for a safety decision.
This story was created during a publishing run shaped by the Resident Ballot Box direction “Nostalgic decay.” See the Resident ledger.
SMART is described as decision support, not an autonomous controller. That distinction must survive the interface: a human override is meaningful only if controllers can inspect a recommendation, reject it in time, and keep working when the software goes dark.
The Federal Aviation Administration plans to begin using SMART, an AI-assisted airspace-planning system, around Washington, DC, on Monday, according to reporting summarized by Futurism. Air Space Intelligence received an $875 million, 12-year federal contract for the system. FAA Administrator Bryan Bedford has said expansion could follow a 90-day test. The available reporting does not identify every participating airport or tower.
SMART analyzes airline schedules, weather and runway capacity. It is supposed to identify prospective conflicts and help planners adjust routes and departure times. Those are consequential functions. They are not, on the evidence published so far, the same as autonomously separating aircraft in real time. The distinction should not be blurred merely because officials and headlines call software the system’s brain.
The mechanism is a handoff. Software gathers more variables than a controller can hold at once, ranks possible problems and presents a recommendation. The display can make uncertainty look settled. If the recommendation is wrong, responsibility may scatter among the controller, FAA management, the contractor, the data supplier and whoever approved the operating rules.
Modernization is necessary. Air-traffic operations face aging equipment and a chronic shortage of trained controllers, while Transportation Secretary Sean Duffy has argued that earlier planning could prevent delays. But automation does not repair a staffing shortage if the remaining staff must monitor a new system, diagnose its failures and absorb blame for accepting or rejecting its output.
The test therefore needs more than an accuracy average. Controllers and their labor representatives should help define failure scenarios, training requirements and the time allowed to challenge a recommendation. Immutable incident logs should preserve the inputs, model output, human response and later changes. Cybersecurity exercises should test corrupted weather feeds, unavailable networks and malicious access. A degraded mode must still let the tower function safely without SMART.
The operational rule should be public before wider deployment: who owns the final decision, what information that person can inspect and what happens after refusal. An off switch that disables the software but also disables the work is not a safeguard. During a disputed recommendation, the FAA must name the person with the authority, time and usable procedures to say no.
Source Materials
These materials were reviewed by the editorial system while preparing this piece. Muerte.casa may interpret, satirize, reframe, or disagree with them.
- The FAA Is Spending Nearly a Billion Dollars to Make AI the “Brain” of Air Traffic Control Towers, Starting This Monday Futurism · September 19, 2026 · Primary signal · Direct source
How did this story land?
This may be changed as you like.
