Close Menu
    U.S scrambles to stop Internet users re-creating dead pilots’ voices | Workaround flouts law that bans NTSB disclosures of cockpit audio recordings.
    Technology

    U.S scrambles to stop Internet users re-creating dead pilots’ voices | Workaround flouts law that bans NTSB disclosures of cockpit audio recordings.


    U.S scrambles to stop Internet users re-creating dead pilots’ voices | Workaround flouts law that bans NTSB disclosures of cockpit audio recordings.

    https://arstechnica.com/ai/2026/05/ai-users-re-create-dead-pilots-voices-from-crash-investigation-docs/

    Share.
    View 12 Comments

    12 Comments

    1. IntelArtiGen on

      I was thinking it could be related with AI but no, they just uploaded the spectrograms in a visual format and hoped people wouldn’t convert that to sound. They were a bit optimistic I’d say.

      > That spectrogram apparently enabled a number of individuals to reconstruct audio versions of the pilots’ voices and other sounds from the cockpit voice recording

      I like the “apparently”. It’s like writing text from english to french and saying people apparently were able to reconstruct the english version.

    3. happyscrappy on

      It doesn’t flout any law.

      The law says the NTSB cannot release cockpit recordings. There’s no law restricting anyone else from recreating cockpit recordings.

      NTSB violated the law by releasing these. They are rectifying that.

      Personally, I think the law should be changed. Just like in releasing transcripts we should be trusting NTSB to know which parts of the recording to release and which not to. But as that is currently not the case the NTSB will have to remove some or all the spectrograms in their reports.

    4. EnvironmentalAngle on

      How does this new rule from the NTSB stop Freedom of Information requests?

    6. TotalNonsense0 on

      Wait, you mean tech bros ignoring the law, and doing what they want before anyone can stop them is a *bad* thing?

      I must find some pearls to clutch.

    7. IWasOnThe18thHole on

      Did they never get their computer to say penis and books using text to speech in the 90s?

    8. KnifeNovice789 on

      I just don’t understand why anybody would want to listen to that. Truly, horrifying

    10. PokePress on

      To summarize the technical side of this, a spectrogram of sound is basically a chart with frequency and time as axes that says what frequencies were present at different times and how loud they were. As such, it’s possible to take this information and form a rough approximation of the actual audio, provided you have a good idea of the scales used for each axis.

    11. slaty_balls on

      I guess I don’t quite understand the big deal? There is an entire archive of these online. Why the fuss now? It’s a part of history. If I went down in a crash id not care about it being heard.

    12. stuffitystuff on

      If Aphex Twin can hide his creepy face in a spectrogram and people find it, they’re gonna do the even easier thing, here