Copyright and AI training

artists are demanding that tech companies stop stealing their work in order to train AI models

OpenAI once argued that it's impossible to train AI without copyrighted materials

If you want your stuff removed, then you have to proactively tell us

we respond swiftly to requests to remove content that rights holders report to us

your responsibility not to use copyrighted materials in your products

a class-action lawsuit, a piracy lawsuit brought by the authors

accusing the company of infringing on their copyright

YouTube videos have been quietly scraped by AI companies

authors who claim their company used their copyrighted creations without permission

they couldn't show harm

the prompt engineer who actually asked for it is the one that would be liable

AI companies may be infringing copyright law by training models on creative works without compensation

The report questioned whether such training qualifies as fair use

OpenAI did admit last year that it would have been impossible to train today's leading AI models without using copyrighted materials.

You can copyright some AI stuff if humans had a hand in creating it, a significant hand.

Merely providing prompts does not meet the threshold.

the lawsuit that these guys filed against OpenAI wasn't for copyright infringement

they can't prove that there's really any damage

expressly prohibit their use in training against artificial intelligence systems

copyright holders can opt out of having their work data mined

They claim the company trained it on a controversial data set that illegally used their books without consent.

That data set includes 196,640 pirated books in it.