Ich habe gerade eine vollständige Aufschlüsselung des bahnbrechenden Gerichtsurteils veröffentlicht, in dem die GEMA (Deutschlands Gesellschaft für Musikrechte) OpenAI verklagte – und gewann.
Im Kern: ChatGPT wurde angeblich auf urheberrechtlich geschützte Songtexte geschult und konnte diese reproduzieren. Das Münchener Gericht entschied, dass es sich hierbei um eine Urheberrechtsverletzung handelte. Es ist das erster großer europäischer Sieg gegen generative KI Verwendung geschützter Inhalte.
Mein Video bricht zusammen:
- Was tatsächlich vor Gericht passiert ist
- Warum die Verteidigung von OpenAI nicht standgehalten hat
- Die globalen Welleneffekte: NYT-Fall, Stabilitäts-KI, Suno und mehr
- Was das fĂĽr Entwickler, KĂĽnstler und KI-Unternehmen in der Zukunft bedeutet
📽️[SehenSiesichhierdievollständigeAufschlüsselungan:[Watchthefullbreakdownhere:https://www.youtube.com/watch?v=dnJ2-3oAy4M
Würde gerne von Bauherren und Juristen hören:
Sollten KI-Unternehmen für Trainingsdaten bezahlen müssen? Oder tötet das Innovationen?
🔍 OpenAI Just Lost a Copyright Case in Germany – Big Win for Creators?
byu/Alive-AI inFuturology
7Â Kommentare
So the issue is the actual end results would have to ‚matter‘. The training is done. It’s over with. If there’s just a fine, it means literally nothing.
I mean, training an AI on non-public data is clearly infringement on a grand scale. I’m baffled that there haven’t been a flood of cases in the US.
AI should only be trained on public domain data, or data that they have a licence to use for (such as wikipedia)
Copyright exists for a reason, and also expires for a reason.
There is no big win no matter what in this regard anymore.
This is no different than meta violating privacy, Google or Microsoft being monopolies, Amazon manipulating prices, etc.
They will pay a joke of a fine equal to a days revenue and then go on to recoup it the following day.
The law has beneath big tech. Period.
I wonder if OpenAI will consider pulling what a few other business have done in the past and just saying we’re taking our ball and going home leaving Germany without access to one of the top ai companies. People are very quickly becoming dependent on this stuff
I think it is rather wild when folks say that “it’s too hard for them to adhere to copyright, just think of what all that would do to AI progress” – like we all owe them our content to progress their shit. lol
Folks act like indeed everything these orgs are creating is out in the open, but reality is there’s plenty they’re protecting, and if you got access to it you’d be sued.
Face it, they’re all training on data they shouldn’t be getting access to without compensating the copyright holder, but we’re letting them off the hook because a) it’s hard, and b) too much $ is wrapped up in building AI to make some folks trillions.
I think preventing training on copyrighted data is a bad idea. Instead they should look into forcing models trained with that data to be free and not profit off of other’s works.