AI is being used to resurrect the voices of dead pilots

NTSB Restricts Docket Access After AI Recreates Pilot Voices from Crash Data

The National Transportation Safety Board (NTSB) temporarily suspended public access to its docket system after discovering that AI tools were used to recreate the voices of pilots killed in a UPS plane crash last year. The incident, involving UPS Flight 2976, has ignited a serious debate over the unforeseen consequences of public data access in an era of powerful generative AI. The agency's swift reaction underscores a new challenge for government bodies: data once considered safe for release can now be transformed into ethically sensitive material by widely available technology.

How a Spectrogram Became Audible Voice

While federal law prohibits the NTSB from releasing raw cockpit audio, the investigation docket for the flight included a spectrogram file of the voice recorder. A spectrogram is a visual representation of sound, mapping frequencies over time. The potential to reverse-engineer this data was first noted publicly by YouTuber Scott Manley. Subsequently, individuals used the spectrogram in conjunction with the official public transcript and AI tools, reportedly including models like Codex, to generate an approximation of the pilots' final communications.

• Source Data: A spectrogram image file of the cockpit voice recorder.
• Enabling Factor: A publicly available transcript providing the text of the conversation.
• Technology Used: Generative AI tools like Codex were reportedly used to synthesize voice audio from the visual data and text.
• Outcome: A reconstructed audio file circulating online, prompting the NTSB to act.

The situation demonstrates a critical vulnerability in legacy data-sharing policies. The NTSB has since restored access to its docket system but kept 42 investigations, including the one for Flight 2976, closed pending a thorough review. This move indicates that the agency is now forced to re-evaluate what constitutes sensitive information and how to properly redact data that could be manipulated by AI, setting a potential precedent for other federal agencies that handle public records.

This incident is a stark illustration of 'data alchemy,' where accessible AI can transmute seemingly inert public data formats into ethically sensitive content. Government agencies and enterprises must now audit their public data not just for what it is, but for what it could become with generative AI, forcing a fundamental rethink of data redaction and release protocols.