GitHub: We going to train on your data after all

https://www.theregister.com/2026/03/26/github_ai_training_policy_changes/

269 points · 22 comments · view on lemmy.world

22 Comments

atzanteol@sh.itjust.works · 76 pts · 137d (4 replies)

It's not GitHub. It's Microsoft. Never forget that.

D1re_W0lf@piefed.social · 12 pts · 137d (3 replies)
  • MicrosoftHub?
  • GitSoft?
  • MicroHub?

🤔

bugfest@lemmy.world · 45 pts · 137d (1 reply)

SlopHub

SleeplessCityLights@programming.dev · 2 pts · 136d

That is the winner right there.

PodPerson@lemmy.zip · 1 pts · 136d

Compuglobalhypermeganet.

MonkCanatella@sh.itjust.works · 28 pts · 136d (1 reply)

I would bet my life savings they already trained on your data

Bieren@lemmy.today · 14 pts · 136d

Right. My theory is that for all of the AI companies that have that don’t use my data option, it’s just an option on the screen. It doesn’t actually do anything. They still use your data.

theywanttocontrolyou@lemmy.world · 15 pts · 137d (1 reply)

Nothing intelligent will come of assimilating my code.

groucho@lemmy.sdf.org · 3 pts · 136d

I think the only thing I have on my github account is an arduino light blinky thing I wrote at 4 AM while crossfaded because a friend needed a glowy wizard staff for a party. It can only make the model worse.

MadMadBunny@lemmy.ca · 12 pts · 137d
Shin@piefed.social · 10 pts · 136d (3 replies)

Stupid, even naive question. What about AGPL code used in this training?
Would that means that the output is also AGPL?

PlexSheep@infosec.pub · 8 pts · 136d

That legal question is not yet clearly answered. I think absolutely yes, but the Megacorps and pro-ai people don't seem to care.

calcopiritus@lemmy.world · 7 pts · 136d (1 reply)

They already trained their AIs on basically all the GPL code. They don't care.

Shin@piefed.social · 1 pts · 136d

Should "we" care?
And if so, what we can do about it?

Greg@lemmy.ca · 4 pts · 136d

lol, enjoy my bugs Octocat

MonkderVierte@lemmy.zip · 1 pts · 135d

Octocat tentacle winks you "bye!".

garbage_world@lemmy.world · -54 pts · 137d (10 replies)
[ removed ]
ugo@feddit.it · 52 pts · 137d (2 replies)
[ removed ]
Venat0r@lemmy.world · 24 pts · 136d

Also the AI is trained on FOSS code but then used to make closed code that ignores the open licenses of the FOSS code that it copied.

Lojcs@piefed.social · 2 pts · 137d

If you use a public ai agent on it your repo isn't really 'private'

MousePotatoDoesStuff@lemmy.world · 8 pts · 136d (6 replies)

> Microslop

> improve

garbage_world@lemmy.world · -5 pts · 136d (5 replies)
[ removed ]
eddiereasoner@mstdn.social · 3 pts · 136d (2 replies)

@garbage_world @MousePotatoDoesStuff hopefully they skip the confirmation bias in the repos with their API keys exposed and commits with a message “hide API key” that we can see the diff of.

garbage_world@lemmy.world · 0 pts · 136d (1 reply)
[ removed ]
eddiereasoner@mstdn.social · 1 pts · 136d

@garbage_world Not sure what the purpose of you comment was, but herego to this path in the repo: secrets/secrets.yaml

Edit: just realized it was encrypted

Also check out apiradar.live

MousePotatoDoesStuff@lemmy.world · 1 pts · 135d (1 reply)

They can pay me for consent to use my data, then.

garbage_world@lemmy.world · -1 pts · 134d
[ removed ]