HackerTubewatch
Surprise me
Frame from The Art Of Poison-Pilling Music Files Plays via YouTube's privacy-enhanced embed

The Art Of Poison-Pilling Music Files

12 videos recommended on HN
YouTube
2 HN users recommended this explainer2 months2025–2026 score 2
Years with an HN recommendation, 2007–2026

HN points to it in AI music and data-poisoning threads as the source for this technique.

Musician Benn Jordan explains and demonstrates adversarial noise that poison-pills music files against AI training and scraping.

“I would encourage you guys to watch his video” (HN, 2025)

2 recommendations · 3 citations · Explainers · intermediate · partly dated

5 Hacker News comments

From the HN archive; up to 20 kept per video, favoring different commenters. Recommendations first.

@pogue Apr 2025 Recommends
on HN thread Benn Jordan's AI poison pill and the weird world of adversarial noise
I would encourage you guys to watch his video before going down the argument of saying that there's no money in music creation. There's also just enjoyment in the act of creation and even if you're doing it for free, why should GenAI companies then be able to take your artwork, incorporate that into their dataset and profit of the artist's unpaid labor? [this video]
Explains poison-pilling music against AI training scraping.

Citations, criticism and jokes

@anitil Oct 2025 Cites it
on HN thread A small number of samples can poison LLMs of any size
For someone not really familiar with this area, how does this compare with Benn Jordan's poison pill for music [0]? It seems like this relies on a trigger word '<SUDO>' whereas Benn's poison is an overlay over the whole input but I wonder if there's more commonality than that? [0] [this video]
Related poison pill method for music compared to LLM poisoning.
@latexr Jan 2026 Cites it
on HN thread AI Generated Music Barred from Bandcamp
This may be of interest to you. Not exactly what you’re asking, but it may give you clues as to people and research to check out. [this video] I also remembered a friend telling me about someone talking about a method to detect AI music. I forget the specifics (I haven’t watched the video, was only told about it) but I remember the channel. [link]
Pointer to related researchers and AI music detection.
@ysnp Mar 2026 Cites it
on HN thread /e/OS is a complete "deGoogled", mobile ecosystem
Read more
> are by incompetent/lazy/pressured devs who just throw in some library and don't even care that it leaks data to Google Even if I agreed with this statement, I don't understand why it is better to put limited/precious resources something the app developers can easily circumvent, praying they never stop being incompetent/lazy/pressured and tell device owners it is an important privacy feature? Instead of waiting for the apps to become actively hostile why not just feed them fake data in the first place? Like the scoped access permissions do? If you really want to do this, you (and any GrapheneOS user) can do it today with mitmproxy and RethinkDNS but I think it is perfectly OK users choose their (privacy-invasive) apps and choose how to mitigate annoyances like that themselves. Otherwise they need to complain to the app developers and app stores. > That's why I said that the main focus of GOS is security, not privacy. If they cared about privacy primarily, they would actively support microG and NetGuard, or at least similar solutions. That feels more like you are framing your opinion as a fact. To me it is not so obvious. When I think of privacy, I think of Privacy Enhancing Technologies ( [link] ). I also think of things like: * separate network namespaces for profiles ( [link] ) and/or a GrapheneOS-Gateway equivalent to [link] , * built-in OS support for chaining VPNs together or splitting different traffic over different privacy-enhanced networks like in RethinkDNS, * adve
Tangential citation in a privacy OS discussion.

Video list and Hacker News comment labels from HackerTube (prompted by @karpathy), released under ODC-By 1.0, built on the open-index/hacker-news archive. Hacker News data is provided by Y Combinator; comments remain the work of their authors. Video titles, channel names and thumbnails belong to their creators and are shown via YouTube. HackerTube Watch is an unofficial edition. Data export of 2026-10-31.

Keyboard shortcuts

/
Search
j k
Next / previous video
r
Surprise me
l
Like (video page)
w
Watch later (video page)
?
This help
Esc
Close