NTSB: UPS Cockpit Voice Recordings Fabricated With AI

Agency blocks public access to accident investigation files after finding unauthorized approximations of CVR.

Meg Godlewski

Friday, May 22, 2026 at 10:29 AM ET

Verified

Edited By:

Travis Tingle

The NTSB says individuals have used AI to animate the crash transcripts. A UPS MD-11F [Credit: Shutterstock/Austin Deppe]

Key Takeaways:

The NTSB discovered individuals are using artificial intelligence (AI) to reconstruct cockpit voice recorder (CVR) audio from sound spectrum imagery released in investigation dockets, including for the 2025 UPS Flight 2976 crash.
This AI reconstruction violates federal law that prohibits the public release of CVR audio due to the highly sensitive nature of cockpit communications.
In response, the NTSB has temporarily taken its docket system offline to examine the scope of the issue and implement solutions to prevent further unauthorized audio reconstructions.

See a mistake? Contact us.

When the National Transportation Safety Board releases an accident investigation docket, it often contains a transcript of the conversations captured on the cockpit voice recorder (CVR)— but never the actual recording.

Yet on Thursday the NTSB announced it had become aware of someone using artificial intelligence (AI) to reconstruct approximations of CVR audio from sound spectrum imagery released as part of the agency’s investigations, including the ongoing probe of the 2025 crash of UPS Flight 2976 in Louisville, Kentucky.