The FAA turned on its SMART AI traffic advisory system nationwide on November 10, 2026, then ordered a rare nationwide ground stop on November 13 over a telecommunications outage that officials did not attribute to the new AI tool. The sequence put two different stories in the same news cycle: a flagship government AI deployment and a safety stop that grounded departing US flights for hours. Separating them matters for anyone building AI that sits next to human operators in regulated environments.
This post walks through what SMART is, what the FAA SMART fact sheet describes, what unions and lawmakers have questioned, and what the November 13 ground stop actually established.
TL;DR
| Question | Answer (as of Nov 13, 2026) |
|---|---|
| What rolled out Nov 10? | SMART expanded from a DC-area pilot to all 22 US air route traffic control centers |
| What does SMART do? | Predicts congestion, suggests routes and timing; humans validate and local leaders accept or decline |
| Reported pilot metrics | FAA cites 180 rerouted flights since Sept 21, ~10 min and ~80 gal fuel saved per flight on average |
| What happened Nov 13? | FAA ATCSCC Advisory 111 ground-stopped all departing traffic for safety |
| Official cause cited | Communications / telecommunications issue at air route centers, per Duffy and AP |
| SMART link? | Denied by ASI; no evidence in major outlet reporting reviewed |
| Contract | ~$875M / 12 years to Air Space Intelligence (FAA and press descriptions) |
What SMART is
SMART is not an autopilot for jets. The FAA's one-page fact sheet describes a cloud-based platform that gives controllers and operators a shared picture of schedules, weather, airport capacity, and airspace conditions before flights depart. It uses AI to predict traffic flows and flag conflicts early enough to adjust routes or timing.
Jeremy Hsu's reporting for Ars Technica quotes the FAA's November 10 statement: after a successful initial rollout in the National Capital Region in September, the agency would use SMART in the airspace of all 22 centers "to reduce delays, reduce fuel consumption and improve the travel experience."
AVweb adds that SMART merges on the order of 200 data streams, including weather, flight paths, volumes, and controller staffing information. FAA specialists in Washington analyze outputs and forward scheduling recommendations to the Air Traffic Control System Command Center in Warrenton, Virginia. Individual facilities then decide whether to implement them.
That human gate is the design point regulators keep repeating. SMART "cannot be used to replace the critical role air traffic controllers play or take over control of an aircraft," the FAA said in its September rollout messaging, as relayed by Ars. Local controllers can accept or ignore each suggestion.
Nationwide expansion and the numbers behind it
The September 21, 2026 pilot focused on congested airspace around Reagan National, Dulles, and Baltimore-Washington. The FAA then extended coverage to centers including Atlanta, Jacksonville, and Miami before the November 10 nationwide switch, according to Ars and Dallas Morning News reporting cited there.
Since the pilot began, the FAA says SMART helped reroute 180 flights, saving an average of about 10 minutes and almost 80 gallons of fuel per flight. Those figures come from the agency's public statements relayed by Ars and AVweb. They are agency-reported operational metrics, not an independent audit, and the FAA has not always named specific flights in press examples.
The underlying contract is large by public-sector AI standards. Ars and AVweb tie SMART to a 12-year, roughly $875 million award to Air Space Intelligence, a Boston company, with additional flow-management modernization work. Our infrastructure coverage often tracks power and siting fights; here the fight is operational trust in software that touches scheduling for the national airspace.
Union and lawmaker skepticism
Airline executives quoted in trade press have voiced support for predictive tools that cut delay and fuel burn. The National Air Traffic Controllers Association (NATCA), which represents more than 15,000 members, told Flying magazine it was not consulted or involved in SMART's "design, testing, or implementation" and therefore could not speak to workforce effects, per Ars.
That gap matters for governance. The FAA positions SMART as decision support, but the people who must live with recommendations did not co-design the system according to NATCA's public statement. Similar tensions show up whenever enterprises roll out AI copilots without operator input; our Strong Safety Evals guide stresses evaluator buy-in for high-stakes deployments.
On October 9, 2026, more than two dozen lawmakers asked the FAA to suspend the initial rollout until questions about safety evaluation, personnel training, and oversight were addressed, AVweb reports. They asked about a planned 90-day evaluation period and standards before wider deployment. The nationwide expansion on November 10 proceeded while those questions remained partly unanswered in public filings we could find.
November 13: nationwide ground stop vs SMART
On the afternoon of November 13, 2026, the FAA issued ATCSCC Advisory 111, ordering all airports to ground stop departing traffic "due to extraordinary circumstances and for reasons of safety" through a window that extended into the next morning UTC. Inbound international flights would be managed case by case.
AP's live coverage and WIRED describe widespread radio frequency and telecommunications outages at major air route traffic control centers, including Boston, Chicago, and New York. Transportation Secretary Sean Duffy said on X that technicians were investigating a communications issue affecting those centers.
Homeland Security Secretary Markwayne Mullin told reporters the administration was not ruling out cyberattack but had no evidence pointing that direction at the time of early briefings, according to Washington Post reporting summarized in search results we cross-checked against AP.
Air Space Intelligence, the SMART vendor, issued a statement that the ground stop was "not based on or a result of the SMART system, any product of ASI or our integration with FAA and ATC," as quoted by the Washington Times. Neither Duffy's early statements nor the FAA advisory text reviewed by explainx.ai linked the halt to SMART.
WIRED explicitly noted the FAA had rolled out SMART nationally earlier in the week and added there was no evidence the ground stop related to that tool. Social speculation conflated the two events because of timing; the verified record as of publication treats them as separate incidents.
What developers are saying
The Hacker News thread on Ars's SMART story was thin on technical detail. gnabgib noted a duplicate submission of the same Ars URL posted minutes earlier (item?id=50490834). No commenter in that thread claimed hands-on SMART experience, which fits a system controllers see but passengers do not.
What this means if you build AI beside operators
SMART is a useful case study for human-in-the-loop AI at national scale:
- Recommend, do not act. The FAA repeats that controllers own separation. Your agent should default the same way in regulated workflows: propose, log, wait for explicit human acceptance.
- Measure the unit of value like AWS does for agents. The FAA cites minutes and fuel per assisted flight, not model accuracy alone. That mirrors Amazon Bedrock's cost-per-task framing for coding agents.
- Staffing and labor are part of the dataset. SMART ingests controller staffing metrics. Ignoring operator capacity in enterprise copilots produces suggestions humans cannot execute.
- Separate failure domains. A nationwide comms outage can halt operations even when an advisory AI stack is healthy. Architect so AI outages cannot take down the primary control path.
- Document evaluation windows before scaling. Lawmakers asked for a 90-day evaluation standard; nationwide rollout outran public answers. Ship with published eval criteria when your product touches safety or compliance.
For broader US public-sector AI deployment patterns, see our posts on New York's AI validation portal and California's AI regulatory freeze fight.
FAA SMART AI nationwide rollout and aviation infrastructure context
Agents that act on your behalf need rules. See how explainx helps secure your agents →
Limits and open questions
We could not independently verify SMART's 180-flight savings math or replay its models on raw FAA data. The $875 million contract figure appears in FAA and press descriptions; we did not pull the full federal procurement file for this post.
It remains unclear how long the November 13 ground stop lasted at each hub, whether a root cause report will cite equipment vendor or cyber factors, and whether lawmakers will reopen their October oversight letter in light of the comms failure days after SMART went national.
SMART's nationwide status also does not answer whether recommendations perform under degraded data conditions. Aviation safety consultant Philip Mann told Ars Technica in September that the next expansion should depend on performance under real workload and degraded data, not demonstration conditions, and on who owns outcomes when a prediction is wrong.
Related reading
- AWS cost per task vs per token — why outcome metrics beat list price for agents
- Anthropic Strong Safety Evals — designing evaluations operators trust
- New York NYCAIRR state AI testing — another government testing model
- California AI freeze lawsuit — state pushback on rapid AI rules
- Google Classroom AI protests — when public AI rollout meets backlash
- FBI ARTEx warning on agentic cyberattacks — federal framing of autonomous tool risk
FAA advisories, contract descriptions, and press attributions are accurate as of November 13, 2026. Ground-stop duration and root-cause findings may change as investigations conclude.