WASHINGTON, D.C. (November 16, 2026) - Two people were injured Saturday night when two planes collided on a taxiway at Miami International Airport. On Friday Secretary of Transportation Sean Duffy called for an additional $50 billion to upgrade the nation's outdated air traffic control [ATC] system, and in June he announced a new system [Strategic Management of Airspace, Routing, and Trajectories (SMART)], costing over $850 million, to help airlines save money by planning for bottlenecks and anticipating schedule conflicts.

But neither is directed, nor likely to, significantly reduce the ever-growing number of near as well as fatal collisions at or near our nation's airports - many going unreported - which could and should have been prevented much less expensively and far more quickly using an existing AI program and an off-the-shelf computer, says Professor John Banzhaf of George Washington University [GWU] where some of the necessary pioneering research has already taken place.

Unfortunately it appears that no current FAA proposal, including ADS-B [Automatic Dependent Surveillance–Broadcast], is designed to make serious use of the tremendous but largely untapped power of AI to prevent deadly airplane collisions and almost weekly near collisions, reports Banzhaf, an MIT-educated engineer with two U.S. patents and considerable experience - including two precedent-setting breakthroughs - with computers.

Earlier this year a Senate committee heard testimony about how two commercial aircraft loaded with passengers almost collided at Boston's Logan Airport, with the two planes coming within a dangerously close 300 feet of each other before the Delta flight performed an emergency evasive maneuver which probably saved the lives of hundreds of passengers.

Just a few months earlier, two fully loaded Southwest jets almost collided over Nashville, and a mass casualty event much larger than the recent one at Washington National was only narrowly averted by the fortuitous reactions of the pilots even before the air traffic controller realized his mistake; and these are only the latest, following an almost weekly number of similar near collisions.

Using AI in different forms to reduce airplane collisions, as well as the escalating number of potentially-fatal near collisions (most of which, says the New York Times, are never reported), was apparently never even evaluated, much less tested, while the FAA under the prior administration reportedly spent some $80 billion promoting diversity rather than improving air traffic safety.

Professor Banzhaf has shown how computers running an existing AI program can help air traffic controllers do what they as humans can't possibly be expected to do by themselves, and to do it tirelessly in milliseconds without becoming distracted, including:
■ monitoring the many different relevant aviation radio frequencies, including commercial, military, government, helicopter, small craft civilian, Homeland Security, ground transportation, and radios from any nearby airports, etc.
■ inputting and including in its analysis the inputs from various radars and video cameras at the airport and on any nearby buildings
■ incorporating in its memory precise maps of the airport including all runways, taxiways, service roads, etc.
■ using all of the above information to determine, in real time and instantaneously, the present position, velocity, and trajectory of all vehicles in the air and on the ground
■ and then calculating various flight paths, trajectories of planes taking off, landing, or taxiing, movements of all ground vehicles, etc.
■ so that it can utilize simple predictive programming to foresee all close calls (near collisions) and possible crashes in fractions of a second
■ and then warn an air traffic controller if there might be a possible problem such as a potential collision, but not itself directly communicate with any aircraft or ground vehicle
■ all while never becoming tired, flustered, distracted, or panicked, or worried by concerns such as a missed paycheck

Other possible solutions to this escalating threat of near collisions, now being proposed by the new administration - including the $50 billion upgrade just announced after the national air traffic shutdown - would require huge expenditures of taxpayer dollars, and take many years to implement; a daily deadly problem exacerbated by the current severe shortage of trained traffic controllers which cannot be remedied quickly.

Fortunately there is something which can be up and running - and be operational at all major airports - much sooner, and at a cost less than to simply update Newark's (or any other individual airport's) outdated equipment (Newark still relies on floppy disks), - argues Professor Banzhaf.

An off-the-shelf already-available desktop computer using a modern AI program could provide a significant increase in air traffic safety; it would be like having a lightning-fast eidetic genius looking over the shoulder of each busy air traffic controller, and providing warnings to the controller of potential collisions in seconds, claims Banzhaf.

While it might not be a complete solution to the current emergency of frequent near misses and some collisions, it would provide a significant important additional margin of safety, at least until longer range solutions can be in place, he says.

If smaller and less complex AI programs can already almost drive cars safely, with all of the many different unusual situations and potential collisions the vehicles must see and react to, they can easily perform the much simpler task of computing flight paths and potential collisions by listening to radio communications in combination with inputs from radar and various cameras, Banzhaf argues.

Indeed, at GWU it has been demonstrated that AI programs can listen to and understand airplane radio communications, and use what it heard to compute flight paths, he says.

More importantly, rather than waiting many months to train new air traffic controllers, and/or spending an additional $50 billion of taxpayer dollars (assuming there are no cost overruns) and many years to update antiquated ATC systems at airports, an off-the-shelf desktop computer running an existing AI program, and monitoring radio traffic, could be ready for proof-of-concept testing much sooner.

Professor Banzhaf's detailed analysis of how such a supplemental warning system built around AI could work, and how it can be tested quickly and safely, has been featured in many reports on the Internet, including, for example:
Let’s Try Using New AI to Reduce Airport Runway Incursions; AI Programs Can Easily Monitor Radio Traffic, Detect Objects, and Analyze

Now that at least one major airport (Newark) recently degenerated into chaos, and very recent collisions (some fatal) occurred at different major airports (with many more narrowly averted), this novel idea should at least be considered and evaluated by other experts before even more lives are needlessly lost, argues Banzhaf.

Fortunately, the GWU professor is no longer alone in calling for testing and then potentially using AI to make flying safer, especially with the FAA's financial support. As just a few examples, consider:

AI Impact Analysis on Aviation Industry
("integration of AI into next generation air navigation systems is enhancing global aviation safety by . . reducing the risk of mid air incidents")

Phoenix, Honeywell Aerospace Test New Tech to Prevent Aircraft Collisions With Ground Vehicles
("it can alert operators of an imminent collision seven to 10 seconds earlier than air traffic control")

Archer’s ZEE AI Predicts Aircraft Movements Minutes Before They Happen
("predicts aircraft movements several minutes into the future, potentially giving pilots and controllers more warning of developing airport ground conflicts")

US Airlines Weigh Safety System to Prevent Midair Collisions Ahead of Legal Mandate, Provider Says
("Some airlines are not waiting for a U.S. legal mandate to look at outfitting planes with an additional system to avert midair collisions")

Aviation Industry Cautiously Optimistic About AI, Finds EASA Survey

Alaska Air CEO Advocates AI Enhancements for Safer, More Sustainable Flights

From clipped communications to clarity: Artificial intelligence is decoding aviation speak for safer flights

ARTIFICIAL INTELLIGENCE (AI) CONTRIBUTION TO AVIATION

Important work already begun at GWU on using AI to enhance airplane safety includes:

A Deep Multi-Agent Reinforcement Learning Approach to Autonomous Separation Assurance

Integrated Arrival and Departure Management for Urban Air Mobility Vertiport Operations in the New York City Airspace

Automated Multimodal Analysis of Air Traffic Around Non-Towered Airports Using Large Language Models

But with all this interest, a compelling life-and-death need, and so much progress recently in using AI in other fields, the Department of Transportation appears to be dragging its feet and risking even more lives by not even considering this new use of AI to improve airline safety, charges Banzhaf, although it is not hesitating to develop its expensive and complex SMART system to help airlines save money, nor to spend some $80 billion promoting diversity.

JOHN F. BANZHAF III, B.S.E.E., J.D., Sc.D.
Professor, George Washington University
FAMRI Dr. William Cahan Distinguished Professor
Fellow, World Technology Network
Inventor of the "Banzhaf Index"
(703) 527-8418
http://banzhaf.net/ jbanzhaf3ATgmail.com @profbanzhaf