The FAA ground stop shows where to test AI-assisted monitoring
Use AI to spot unusual delays or gaps in communications, compare incident reports and prepare recovery options. Test the backup and keep operating decisions with authorised people.
The outage and the recovery
On 13 November, the US Federal Aviation Administration (FAA) paused departures nationwide after a telecommunications outage affected air traffic control centres. It lifted the stop that evening. Officials described an overload in an older system during an upgrade and reported no indication of a cyberattack. By 7:05 p.m. Eastern, Cirium had counted 827 diversions, including flights that turned back, according to AP. Airlines still had to reposition aircraft and crews after the stop ended.
FAA ground-stop order Associated Press on the outage and recovery Associated Press live reporting
What officials said about the FAA’s AI tool
The FAA had announced a nationwide expansion of SMART, an AI-supported tool that helps specialists recommend traffic changes. Homeland Security Secretary Markwayne Mullin said investigators would examine a possible link to the outage. He reported no indication of a link at the time. The supplier, Air Space Intelligence, said its system did not cause the ground stop. The sequence that led to the outage remained under investigation Friday evening.
FAA notice on the SMART expansion Associated Press on the outage and recovery Associated Press live reporting
Use AI to detect and organise
First, measure how often messages arrive late or go missing across sites during normal operations. A model could flag a departure from that baseline. Show the operator which signals changed, the time of the last valid reading and whether the monitoring feed itself is stale. During an outage, a separate assistant could group duplicate tickets and draft a timeline from approved logs. Require a source and timestamp beside each entry. Human incident commanders own escalation and recovery decisions. Give the assistant no permission to issue operational instructions or clear aircraft.
Test the backup without the primary service
In a tabletop exercise, simulate missing monitoring data and contradictory supplier updates. Separately, switch off the primary connection in a controlled technical test. Check whether both paths depend on the same carrier or cloud region. Compare AI-assisted detection with a simple threshold alert on the same recorded data. Measure time to a verified alert, missed incidents, false alarms per shift and time to restore an agreed minimum service. Keep the option that gives operators the clearest evidence under pressure. NIST’s AI Risk Management Framework provides a structure for documenting those controls and reviewing the results.
Ask for evidence before the next upgrade
The US Government Accountability Office documented longstanding problems with ageing air traffic control systems, and the US Department of Transportation’s upgrade plan includes changes to telecommunications and radios. The November outage makes the transition a risk to manage in its own right. Ask the system owner for a witnessed fallback test, a rollback threshold and a named person who can authorise restart. Keep the same requirements when the proposed upgrade includes AI.
US Government Accountability Office on ageing air traffic control systems US Department of Transportation modernisation plan