Looking at a timeline of cases against various AI companies suggests that’s not quite the case. This page had a good overview, showing how cases are being resolved. Some of the recent notable outcomes involve the German courts finding OpenAI violated copyright laws, OpenAI being forced to reveal internal communications about trying to hide a massive dataset of pirated books, and a class action suceeding against Anthropic, but there’s a bunch more.
It’s important to separate the training AI part from the conventional copyright violation. A lot of these companies downloaded stuff they shouldn’t have downloaded and that is a copyright violation in its own right. But the training part has been ruled as fair use in a few prominent cases already, such as the Anthropic one.
Beyond even that, there are generative AIs that were trained entirely on material that the trainer owned the license to outright - Adobe’s “Firefly” model, for example.
So I have yet to see it established that generative AI inherently involves “asset theft.” You’ll have to give me something specific. That page has far too many cases jumbled together covering a whole range of related subjects, some of them not even directly AI-related (I notice one of the first ones in the list is “A federal judge accused a third-party law firm of attempting to “trick” authors out of their record $1.5 billion copyright class action settlement with Anthropic.” That’s just routine legal shenanigans).
Looking at a timeline of cases against various AI companies suggests that’s not quite the case. This page had a good overview, showing how cases are being resolved. Some of the recent notable outcomes involve the German courts finding OpenAI violated copyright laws, OpenAI being forced to reveal internal communications about trying to hide a massive dataset of pirated books, and a class action suceeding against Anthropic, but there’s a bunch more.
It’s important to separate the training AI part from the conventional copyright violation. A lot of these companies downloaded stuff they shouldn’t have downloaded and that is a copyright violation in its own right. But the training part has been ruled as fair use in a few prominent cases already, such as the Anthropic one.
Beyond even that, there are generative AIs that were trained entirely on material that the trainer owned the license to outright - Adobe’s “Firefly” model, for example.
So I have yet to see it established that generative AI inherently involves “asset theft.” You’ll have to give me something specific. That page has far too many cases jumbled together covering a whole range of related subjects, some of them not even directly AI-related (I notice one of the first ones in the list is “A federal judge accused a third-party law firm of attempting to “trick” authors out of their record $1.5 billion copyright class action settlement with Anthropic.” That’s just routine legal shenanigans).