FAA uses – Facing an overwhelming flood of flight and incident data, the FAA is expanding AI tools to help human analysts spot patterns, validate findings, and track leading indicators before problems become headline events.
The Federal Aviation Administration’s safety staff doesn’t just watch for disasters—they try to make sense of the massive data trail left behind by every flight.
But the FAA has a data problem: every flight generates so much information that analysts struggle to sift through it all and turn it into clear trends. The agency’s leaders argue the goal isn’t to stop responding to incidents—it’s to see problems earlier. using artificial intelligence to help humans interpret the warning signs that come before something goes wrong.
That framing has urgency because the aviation world is living through a streak of high travel volumes and frequent. high-profile accidents and close calls. In February. National Transportation Safety Board Chair Jennifer Homendy criticized the FAA during a hearing before the Senate Commerce. Science and Transportation committee for ignoring statistics that later surfaced in accident investigations. Former Department of Transportation Inspector General Mary Schiavo also described the FAA as the “tombstone agency. ” a reference to how investigations often crystallize only after harm.
Within the FAA. Jodi Baker. the deputy associate administrator for aviation safety management. put the challenge in blunt terms: “A bad thing happening is a lagging indicator because the bad thing already happened.” She said the agency has “lots and lots and lots of data sources. ” and the long-standing question has been how to “glean intelligence out of data sources.”.
Baker said new AI technologies can help the FAA analyze more efficiently and strengthen safety across the nation’s airspace—while staying firmly in the hands of human decision-makers.
The FAA is using AI in ways it can control—streamlining analysis. consolidating signals. and improving how qualitative reports are understood. In the aftermath of the DCA crash in 2025. the agency conducted an AI analysis to help determine where mixed helicopter traffic near other airports should be limited. The move aligned with a recommendation from the National Transportation Safety Board in the wake of the crash.
The FAA has also applied AI to incident reports, combing through them regularly. During the government shutdown in the fall of 2025, it conducted an AI analysis intended to help optimize flight schedules.
Even as those examples show AI taking on more of the heavy lifting, Baker stressed that the agency draws a strict line between assistance and autonomy. “AI is not running amok,” she said, describing how models are trained on the FAA’s data systems and do not “roam the internet.”
“People are involved with validating what AI is telling us,” Baker said. “AI for us is a decision-making tool, but it is not a decision maker. So, people are still involved with the system.”
One application Baker described as especially helpful is AI’s ability to work through narrative incident reports—where information is often buried in text and naming conventions vary. She said she has spent years in conversations where one database calls a plane a Boeing 737. another lists it as a B737. and another writes it as a B-737. That kind of messy reality, she said, forces humans to “marry” entries together and spend time cleaning.
With AI trained to recognize that variation, Baker said the system can pull the information together quickly: “you can train AI to say all of these things are a Boeing 737. And so, it can pull it all together really quickly.”
As FAA AI models are trained at the agency. they are expected to incorporate more parameters and flag potentially dangerous situations more accurately. Baker also said AI could be used to test the outcomes of rule changes before implementation—an approach aimed at analyzing effects preemptively—and to optimize resource allocation to address issues.
In Baker’s view, the most important shift is timing. “It’s a real-time ability to see what’s going on in the system,” she said. “In the past, you’d wait.”
She described a problem many safety teams face: without strong tools, leading indicators are difficult to track. “I didn’t really have a good way of tracking leading indicators before,” Baker said. “And now we can track leading indicators. we can track the system as it’s operating and see what’s going on.”.
That push toward earlier visibility isn’t limited to the FAA. Across the industry, researchers and aviation safety experts say AI’s value lies in turning large volumes of information into usable baselines—then flagging what falls outside “normal.”
Kristy Kiernan. associate director for the Boeing Center for Aviation and Aerospace Safety at Embry-Riddle Aeronautical University. said AI can be “very helpful in taking a huge amount of data and learning from it what is normal. ” including “external factors.” From there. AI can “flag the outliers from normal behavior” so humans can examine why those deviations occur.
Kiernan also argued that AI could improve safety by giving stakeholders more time and capacity to analyze positive trends. not just respond to dangerous incidents. “The vast majority of our flights end safely. and we’re not looking at them because they end well. ” she said. “Don’t you want to be looking at the stuff you do right?”.
She said the point is to focus on what goes right and understand why: “We want more good things to happen, we want to understand why things are going right.”
Even with that promise, Kiernan warned that AI has limitations and must remain part of a broader safety system. She compared aviation safety work to cockpit teamwork. where pilots receive training in human factors to understand both their strengths and limitations. including those of their human copilot.
“Where you run into problems is not because of the weaknesses specifically, every system has them, it’s when you’re unaware of it,” Kiernan said. “Calibrating people’s trust effectively – you don’t want to over-trust AI.”
AI, in other words, can’t be treated like a substitute for training, judgment, or proven safety practices.
That caution extends to the question many readers ask when they see AI entering safety: will it replace pilots?. Kiernan said it won’t, and she pointed to a blind spot in how accident statistics are sometimes presented. “You’ll read a lot that 80% of accidents are caused by human error,” she said. “What that does not consider is: how many times has a human intervened in a situation to keep it safe. which was never captured in a data set. If you change the role of a human operator without accounting for the positive role they have been contributing. you introduce uncertainty.”.
Baker’s reaction to the FAA’s expanding use of AI was different in tone, but the message was similar: she sees AI as a new tool for safety analysts, not a break from the human work that underpins aviation.
“I’ve been working in safety for 20 plus years, and this is just a really exciting time,” she said. “I mean,” she continued, framing it as a shift toward seeing what’s happening earlier—before the industry only has the clarity that comes after something bad already occurred.
For the FAA, the next chapter in safety analysis hinges on a simple but demanding promise: let AI help the agency find the patterns that humans may not have time to surface, then keep people in charge of what those patterns mean.
FAA artificial intelligence flight safety aviation safety flight data analysis incident reports predictive analysis NTSB Jennifer Homendy Jodi Baker aviation technology