The federal government just cut an $875 million check for an AI-powered air traffic management system while the infrastructure it runs on is so fragile that a construction crew with a shovel grounded hundreds of flights across the Northeast on Monday.

The contradiction is the whole story. Washington wants artificial intelligence for itself — and wants to regulate it out of your hands. The same day Transportation Secretary Sean Duffy debuted the new "SMART" tool at DC-area airports, a severed fiber line from an Amtrak work crew crippled the Philadelphia Terminal Radar Approach Control facility, forcing ground stops at Newark, Philadelphia, Teterboro, LaGuardia, and JFK. Over 500 flights canceled. Hundreds more delayed. Pilots literally asking controllers what was going on and getting "no clue" for an answer, according to air traffic audio obtained by NBC News.

Duffy framed the rollout as a leap forward. The Strategic Management of Airspace, Routes and Trajectories tool puts 200 data streams on a unified platform and uses AI to predict weather disruptions so managers can reroute flights before snarls form. "You're going to be on the airfield much faster. Your airline is going to use less fuel," Duffy told reporters. He was quick to insist: "We are not outsourcing aviation to AI." He called it "dumb AI" — an advisory tool, not a safety system. Humans still make the calls.

Fine. But follow the money. Boston-based Air Space Intelligence landed an $875 million contract over 12 years for SMART and a similar system, bankrolled by the $12.5 billion in the One Big Beautiful Bill Act earmarked for air traffic infrastructure overhauls. One company. Nearly a billion dollars. No bid details, no competing small businesses in sight — just the usual conveyor belt from federal coffers to a single vendor.

Meanwhile, the system SMART sits on top of is rotting. The Guardian highlighted a Government Accountability Office report published the same day as the meltdown, warning that the FAA has not finished risk assessments for spoofing, jamming, and other attacks on the radio spectrum. Two systems used to send text messages to aircraft predate modern cybersecurity standards and lack encryption. The FAA has no real-time system to detect these threats — it can only investigate after the fact. Senator Ron Wyden called it "a major threat to US national security, to our economy and the safety of the flying public."

NBC News noted this is the same corridor where an equipment malfunction last year left roughly a dozen planes flying blind for 90 seconds. The Post buried the GAO's cybersecurity warnings entirely, focusing on SMART's promised efficiencies. The Guardian, by contrast, led with the hacking threat and treated the SMART rollout as a footnote.

So here is where things stand: The feds are spending nearly a billion dollars on AI to optimize flight routes on a system where a cut fiber line can paralyze the busiest airspace in the country, where controllers still rely on old-fashioned paper maps, and where a hacker could theoretically send false clearance cancellations to aircraft with no encryption standing in the way. And the same government building this AI-powered future is the one drafting regulations to make sure you can't access comparable tools without its permission.

The question isn't whether AI belongs in air traffic management. It's who gets to use it — and who decides.