Without a specific chain of custody for every piece of training data going into the models, there is a default that the model cannot be trusted and is likely infringing on someone’s copyright.
To specify the 'nothing AI is free" part, LLMs are grossly computationally inefficient. Whether it’s local or not.
Those are the settled ones so far. This is 4 years into AI existing. Lawsuits, especially copyright lawsuits, tend to take up to a decade in the US, because the US legal system is shit.
In February 2025, the court granted summary judgment that such copying was not fair use, emphasizing that the purpose of ROSS’s copying was to build a directly competing product
Bartz v. Anthropic
The court granted summary judgment for Anthropic that training LLMs on copyrighted materials is fair use.
Kadrey v. Meta
Judge Vince Chhabria denied plaintiffs’ motion for partial summary judgment on fair use and granted Meta’s cross-motion and granted Meta’s motion for partial summary judgment on the DMCA claim.
UMG v Udio
Settled - no judgement.
Those are the settled ones so far. This is 4 years into AI existing. Lawsuits, especially copyright lawsuits, tend to take up to a decade in the US, because the US legal system is shit.
Here’s 118 currently in progress. Because AI is copyright infringement.
Having lawsuits is not winning lawsuits. The AI companies have been winning on fair use. Same as Google back in the 2000s when they were sued for for various search products (news, books, etc.).
Hi this must be your first time on Earth in the last decade, every single AI company has been in or is currently in no less than ten dozen lawsuits over copyright infringement. It’s so bad there’s at least one website purpose built to track copyright infringement from AI companies..
Without a specific chain of custody for every piece of training data going into the models, there is a default that the model cannot be trusted and is likely infringing on someone’s copyright.
To specify the 'nothing AI is free" part, LLMs are grossly computationally inefficient. Whether it’s local or not.
Already installed on most distros.
So where are the winning lawsuits of all that copyright infringement then?
And they are called… ?
Thomson Reuters v. Ross Intelligence
Bartz v. Anthropic
Kadrey v. Meta
UMG v Udio
Those are the settled ones so far. This is 4 years into AI existing. Lawsuits, especially copyright lawsuits, tend to take up to a decade in the US, because the US legal system is shit.
Here’s 118 currently in progress. Because AI is copyright infringement.
In February 2025, the court granted summary judgment that such copying was not fair use, emphasizing that the purpose of ROSS’s copying was to build a directly competing product
The court granted summary judgment for Anthropic that training LLMs on copyrighted materials is fair use.
Judge Vince Chhabria denied plaintiffs’ motion for partial summary judgment on fair use and granted Meta’s cross-motion and granted Meta’s motion for partial summary judgment on the DMCA claim.
Settled - no judgement.
Having lawsuits is not winning lawsuits. The AI companies have been winning on fair use. Same as Google back in the 2000s when they were sued for for various search products (news, books, etc.).