Twenty-three Democratic members of the U.S. House Committee on Transportation and Infrastructure have asked the Government Accountability Office to review a new artificial-intelligence-driven air traffic management tool the Federal Aviation Administration is rolling out.

In an Oct. 13 letter viewed by The Air Current, the lawmakers cited two concerns: the tool’s reliability, and its potential to “hallucinate safety-critical information.”

What does the letter actually request?

The members want the GAO — Congress’s watchdog arm — to examine the FAA’s new system as it enters operational use. Their worry is straightforward: an AI tool that generates false or fabricated outputs in a safety-critical environment could put bad information in front of controllers.

The phrase at the center of the letter, “hallucinate safety-critical information,” borrows from the vocabulary of large language models, where a system confidently produces content that is simply wrong. Applied to air traffic management, the lawmakers are asking what happens if such a failure occurs during live operations.

Who is involved?

• The FAA, which is deploying the AI-driven air traffic management tool.
• The GAO, asked to conduct the review.
• 23 House Democrats on the Committee on Transportation and Infrastructure, signatories of the Oct. 13 letter.
• The Air Current, which viewed the correspondence.

Why it matters

The request arrives as the FAA pushes automation into airspace management, a domain where regulators have historically required deterministic, certifiable systems. A GAO review would put an independent auditor between the FAA’s deployment schedule and Congress’s oversight of it.

For airlines and controllers, the stakes are operational: any tool that shapes traffic flow decisions sits inside the chain that determines separation, routing and capacity. The lawmakers’ letter signals that at least part of the House transportation committee wants verification before trust.

What the GAO decides to examine — and whether the FAA pauses any part of the rollout while the review proceeds — will shape how quickly AI reaches the working level of U.S. air traffic control.

By Nathan Brooks