AI users re-create dead pilots’ voices from crash investigation docs

In a striking development, internet enthusiasts have harnessed AI and software tools to recreate the voices of pilots in the final moments of a tragic cargo plane crash. This emergence of reconstructed audio has led the U.S. National Transportation Safety Board (NTSB) to temporarily halt public access to its database of civil transportation accident information, as federal regulations restrict the release of cockpit voice recordings. Typically, the NTSB provides factual reports and investigative findings related to aircraft incidents. However, on May 21, the agency announced that its online docket system would be 'temporarily unavailable' while they assess the publicly accessible materials that allowed individuals to reproduce cockpit audio from past accidents. The NTSB acknowledged the advancements in image recognition and computational techniques, which have enabled the reconstruction of cockpit voice recorder audio from sound spectrum imagery released during investigations. This includes the ongoing inquiry into the UPS flight 2976 crash in Louisville, Kentucky, which occurred on November 4, 2025. The crash resulted from a structural failure that caused an engine to detach shortly after takeoff, tragically claiming the lives of all three pilots and an additional 12 individuals on the ground, while injuring 23 others. Federal law, enacted by Congress in 1990, explicitly prohibits the NTSB from making any part of cockpit voice or video recordings public, a measure aimed at safeguarding the privacy of aircrew members. This legislation was introduced following a backlash from airline pilots against a TV station's airing of cockpit conversations related to the Delta Air Lines Flight 1141 crash in 1988. As technology evolves, the implications of recreating such sensitive audio continue to spark debate and concern within the aviation community.