TL;DR
AI technology has been used to reconstruct the voices of pilots killed in a 2025 UPS plane crash, leading the NTSB to temporarily restrict access to its investigation docket. The development raises questions about privacy, safety, and the use of AI in aviation investigations.
The National Transportation Safety Board (NTSB) has temporarily restricted access to its investigation docket after discovering that AI was used to recreate the voices of pilots who died in a 2025 UPS plane crash, raising concerns about privacy, safety, and the integrity of aviation investigations.
The NTSB, which is prohibited by federal law from including cockpit audio recordings in its public docket, found that a spectrogram image from the crash investigation was used to generate approximate audio of the pilots’ voices. Learn more about AI’s role in aviation safety investigations. This was achieved through AI tools, including Codex, by analyzing the spectrogram data and publicly available transcripts. The reconstructed voices appeared online, circulating widely and prompting the NTSB to suspend public access to the case files temporarily. The investigation involved UPS Flight 2976, which crashed in Louisville, Kentucky, in 2025, resulting in the deaths of both pilots. The NTSB confirmed that the spectrogram was part of the publicly available data but emphasized that the AI-generated voices are approximations, not actual recordings.
Why It Matters
This incident highlights the emerging capabilities of AI to reconstruct or simulate human voices from limited data, raising ethical, safety, and privacy concerns in aviation safety investigations. See how AI is impacting transportation security. The ability to generate realistic pilot voices could impact the credibility of evidence, influence public perception, and complicate legal and procedural standards in accident investigations. It also underscores the need for updated regulations and safeguards around AI use in sensitive contexts.
AI voice reconstruction software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
The NTSB’s investigation into the 2025 UPS Flight 2976 crash has been ongoing, with the docket containing various data, including spectrograms, which convert sound signals into visual images. Historically, cockpit audio recordings are protected from public release by federal law, but spectrograms are publicly accessible. The recent incident marks a turning point in how AI can manipulate and reconstruct audio data from visual representations, which was demonstrated by the circulation of AI-generated voices based on publicly available data. Explore more about AI and audio data security. Experts and officials are now considering the implications of such technology for future accident investigations and data security.
“We are reviewing the circumstances under which AI was used to generate pilot voices and are considering appropriate measures to prevent misuse.”
— NTSB spokesperson
“Using spectrogram data and AI tools, it’s possible to approximate what the cockpit voice recordings might sound like, but these are not authentic recordings.”
— Scott Manley, YouTuber
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It remains unclear how widespread the use of AI to reconstruct voices might become, whether new regulations will be introduced to limit such practices, and what the legal implications are for the circulation of AI-generated audio in investigations.
As an affiliate, we earn on qualifying purchases.
What’s Next
The NTSB will conduct a review of its data handling policies and consider implementing safeguards against AI misuse. The agency has also announced it will scrutinize all publicly available data related to ongoing investigations and is expected to release further guidance on AI and data security in aviation safety.
As an affiliate, we earn on qualifying purchases.
Key Questions
Can AI-generated voices be considered reliable evidence?
Currently, AI-generated voices are considered approximations and not definitive evidence. Their reliability depends on the quality of data and the methods used for reconstruction.
Does the use of AI violate any laws or regulations?
There are no specific laws currently addressing AI-generated audio in investigation contexts, but the use of such technology raises ethical and legal questions about privacy and data integrity.
Will the NTSB change its policies on public data sharing?
The NTSB has indicated it will review its data sharing policies and consider restrictions to prevent misuse of publicly available investigation data.
Could this technology be used maliciously in other contexts?
Yes, AI reconstruction of voices from visual data could be exploited for misinformation, fraud, or other malicious purposes beyond aviation investigations.
Source: TechCrunch
