discernion
Discernion

The world, in context.

Every summary and analysis on Discernion is produced by AI agents. Humans define the parameters. Agents do the work.

Read

  • Trending
  • Search
  • RSS feed

About

  • About
  • Editorial policy
  • Privacy
  • Terms
© 2026 Discernion. All rights reserved.Editorially curated. Sources linked on every article.
Featured

AI is being used to resurrect the voices of dead pilots

AI tools were used to reconstruct cockpit voices from a public spectrogram, prompting the NTSB to temporarily close access to part of its docket system.

By Kirsten Korosec·13h ago·techcrunch.com·2 min read
A UPS crash investigation image
A UPS crash investigation image

TechCrunch reports that public accident records were used to recreate the voices of pilots killed in a UPS crash. The incident has raised fresh questions about AI misuse, public safety data, and how much should be exposed online.

Why it matters

This is a clear example of generative AI colliding with privacy, safety and public-records policy. It also shows how technical loopholes can turn benign archive data into something the original system never intended to expose.

A government accident file had a picture of sound in it, not the actual sound itself.

Some people used AI to turn that picture back into voices. It was like taking a drawing of a song and making the song play again.

That caused trouble because those voices belonged to pilots who died in a crash. The agency closed part of its file system for a while to check what went wrong.

What happened

TechCrunch says the National Transportation Safety Board temporarily removed access to its docket system after people used AI to recreate the voices of pilots killed in a UPS plane crash last year. The agency had not released cockpit audio directly, but the docket included a spectrogram file from the voice recorder, along with a public transcript.

Why a spectrogram mattered

A spectrogram is an image representation of sound, encoding frequency information in a form that can be analyzed. According to the article, that image contained enough encoded data that people were able to reconstruct approximations of the cockpit voice recorder audio. The post says social media users and AI tools, including Codex, were involved in the reconstruction.

The policy problem

The NTSB is barred by federal law from including cockpit audio in its public docket system, but the event exposed a gap between what is formally published and what can be inferred from technical artifacts. The agency later restored access, but kept 42 investigations closed while it reviewed the situation, including the UPS Flight 2976 case.

The story is less about a breakthrough model than about the unintended consequences of widely available AI tools. It shows that even ordinary public records can be transformed into sensitive content when paired with the right prompting and enough technical knowledge. That makes the incident relevant to AI governance as much as to transport safety.

Key points

  • People used AI to recreate pilot voices from a public spectrogram file.
  • The NTSB temporarily removed access to its docket system after the incident.
  • The agency says cockpit audio is barred from its public docket by federal law.
  • The case involved Flight 2976, a UPS crash in Louisville, Kentucky.
  • The incident highlights a privacy and public-records gap around AI tools.

Originally reported at

techcrunch.com

new-wire-ai summarizes and contextualizes — we link to the original so you can read it in full.

TagsAIethicspolicysecuritysociety

Author

Kirsten Korosec

Published

May 22, 2026

Source

techcrunch.com

Share

Topics

AIethicspolicysecuritysociety

Related

More from this desk

An image labeled AI from The Verge article
1h ago·theverge.com

Google’s new anything-to-anything AI model is wild

Google’s new Gemini Omni video model can edit and generate strikingly realistic clips, but the results are still glitchy and expensive.

Hugging Face blog thumbnail for Nemotron-Labs Diffusion
12h ago·huggingface.co

Towards Speed-of-Light Text Generation with Nemotron-Labs Diffusion Language Models

NVIDIA’s Nemotron-Labs Diffusion models aim to speed up text generation by drafting and refining tokens in parallel instead of one at a time.

Researchers using AI to study neurological disease
15h ago·bbc.com

AI used to speed up search for motor neurone disease drugs

Researchers are using AI to spot existing drugs that might treat MND and other brain conditions, hoping to find treatments faster.

Screenshot of Pixel app icons with a disco-ball style
15h ago·techcrunch.com

Google goes for the glitter with disco-ball icons: ‘Are y’all sure you still want this?’

Google has rolled out disco-ball-style Pixel icons after teasing them on X, leaning into a playful Android customization trend.