The US Department of Transportation is putting AI in the loop on air traffic management, and the price tag is $12 billion. Transportation Secretary Sean Duffy confirmed the plan at a Semafor event last week, then went on CBS News Tuesday with more details, according to Futurism AI. The program is called SMART (Strategic Management of Airspace Routing Trajectories), and three vendors are now competing for the contract: Palantir, Thales SA, and Air Space Intelligence.
Palantir already told investors it had been contracted by the FAA to “provide a data analytics tool that will help advance the agency’s modernization objectives for aviation safety.” Bloomberg got the full vendor list through a person familiar with the matter. Duffy, who briefly ran NASA before this gig, says the software will help controllers schedule flights weeks ahead to cut down on delays.
Here’s how he framed it on CBS: “This software will say, ‘well, listen, we can see this 45 days out. Let’s move some of those flights a little bit later, or five, seven, 10 minutes earlier, and we can resolve the issue. And so then you are not delayed.'” Duffy added that SMART will stop short of “replacing humans in how we manage the airspace.”
What’s actually changing
- Predictive scheduling stretches out 45 days, instead of reactive day-of juggling.
- Three private vendors compete, with Palantir already publicly confirmed.
- Humans stay in the loop on live airspace control, per Duffy.
- $12 billion is the headline cost taxpayers will foot.
Why this matters
The FAA has been chronically understaffed and running on infrastructure that dates back to the 1980s. Modernization talks have come and gone for two decades with little to show. SMART is the first concrete bet that machine learning can handle the part controllers find hardest: forward-looking scheduling under uncertainty. If it works, delays drop and controllers focus on live traffic. If it doesn’t, the failure mode lands on millions of passengers.
Futurism AI raises the obvious concern, and it’s a fair one. AI systems hallucinate. They’ve stumbled at staffing, traffic predictions, and even running vending machines. The publication’s line is sharp: if today’s best AI can’t run an office snack machine without effectively forming a cartel, putting it in charge of national flight scheduling is a leap of faith.
The vendor angle
What stands out is the lineup. Palantir brings defense-grade data infrastructure and a track record with federal contracts. Thales SA has been embedded in aviation systems for decades, mostly on the hardware and avionics side. Air Space Intelligence is the smallest of the three and the most ML-native. Whoever wins SMART will set the template for what “AI in aviation” looks like for the next decade.
What practitioners should watch
- Whether the FAA publishes accuracy benchmarks for SMART before deployment.
- How human-in-the-loop is actually implemented: advisory recommendations or binding decisions.
- Whether vendors release model cards, or keep the system fully black-box.
- Public reporting on near-misses or scheduling errors tied to AI recommendations.
The bigger trend is unmistakable. Federal agencies are no longer piloting AI on the edges. They’re wiring it into critical infrastructure. SMART joins the IRS, DoD, and HHS in moving AI from experiment to operations. The contracts are bigger, the stakes are higher, and the accountability layer is still being figured out as the systems go live.
More details on the rollout, the vendor competition, and the broader concerns about AI reliability are available at the original Futurism AI report.