Elektrine lite

← Feed

@lina@vt.social

Post #4067226

2026-07-24 14:31 UTC

Looks like Hugging Face scraped ~all of GitHub into an AI training dataset... https://huggingface.co/spaces/HuggingFaceCode/in-the-stack There is an opt-out link. You might want to use it.

Replies (4)

  • @iamada@tech.lgbt 2026-07-24 14:36

    @lina@vt.social amazing how if any of us did that, we’d get arrested and prosecuted under anti piracy laws

    Open ##4083515

  • @codecat@meow.social 2026-07-24 15:02

    @lina@vt.social of course the opt out process requires making a github issue. ugh. open source sucks now.

    Open ##4205616

  • @Chris@fedi.chris-macleod.ca 2026-07-24 16:21

    Yes, there is code from 8 repositories in The Stack v3 Wow, even my shitcode is included. I may be the odd one out in that I don't care if what I make is used for training, but I almost feel sorry for anyone using a model that learned from my casual and student projects 😆. They say they exclude files identified as non_permissive, but it's not clear from the description alone what they're doing with files identified as no_license. At a glance, they seem to be included by default. At any rate, they've clearly got stuff in there that isn't permissively licensed.

    Open ##4205618

  • @cinebox@masto.hackers.town 2026-07-24 17:44

    @lina@vt.social There are still opt-out requests open from 2024. I guess they decided it was too inconvenient to be told “no”

    Open ##4249574