Data Poisoning could be a tool we use to identify AI that has used copyritten material
2026-05-16 14:08 UTC
Data Poisoning could be a tool we use to identify AI that has used copyritten material, or we use it to mess with AI.
vice.com/…/infinite-ai-homer-simpson-cover-songs-…
djmag.com/…/spotify-and-major-labels-sue-shadow-l…
mosis.eecs.utk.edu/harmonycloak.html
mosis.eecs.utk.edu/…/meerza2024harmonycloak.pdf
www.eff.org/cyberspace-independence
Replies (4)
-
@cypherpunks@lemmy.ml 2026-05-16 14:39
> identify AI that has used copyrighted material but, that is basically all modern "AI". (the only LLM i've heard of which actually claims that its training corpus is freely licensed is [Apertus](https://en.wikipedia.org/wiki/Apertus_(LLM))...)
-
@very_well_lost@lemmy.world 2026-05-16 15:07
People have actually been doing this to catch plagiarism for centuries, long before LLMs were a thing. See [trap streets](https://en.wikipedia.org/wiki/Trap_street) for one of the better known examples.
-
@sidefaceturdtalker@leminal.space 2026-05-16 22:14
One way to push back forsure but we need to refresh the tree of liberty asap
-
@ExtremeDullard@piefed.social 2026-05-16 18:42
This idea is [as old as books](https://en.wikipedia.org/wiki/Fictitious_entry).