> The lawsuit cites specific ChatGPT searches for each author, such as one for Martin that alleges the program generated “an infringing, unauthorized, and detailed outline for a prequel” to “A Game of Thrones” that was titled “A Dawn of Direwolves” and used “the same characters from Martin’s existing books in the series “A Song of Ice and Fire.”
Selling this work would be copyright infringement. There’s no ambiguity there.
The gpt copyright infringement is strangely divided a lot of tech workers.
Generally people that are against overly broad patents and other copyrights should be dismissive of copyright claims of artists complaining about gpt. I remember Oracle being mocked for suing google over 9 lines of fairly generic code. Or Tiffany trademarking a color.
But now because Microsoft stands to benefit people are very concerned about "stolen" work. The funny thing is that the overall net benefit would be a lot greater from LLMs as opposed to someone not being able to use a certain color.
The biggest irony is that openai will obv prevail regardless. They'll pay their fines and work with regulators. The only thing that will be hurt will be competitors or open source projects
It's just annoying that after decades of copyright-maximalist public policy and corporate PR, it's now time for a sensible application of copyright when it comes to this specific issue.
> selling this work would be copyright infringement. There’s no ambiguity there.
>Generating it would not.
Are you sure about that? My understanding is that under US copyright law the copyright owner has the exclusive right "to prepare derivative works based upon the copyrighted work."
Which is to say, a derivative work - which this would certainly be - would legally still be infringing even if it were not distributed.
It’s fair use though. You would have an extremely difficult time arguing that a piece of generated art for just yourself passes muster on:
> the effect of the use upon the potential market for or value of the copyrighted work.
Because the effect of the use on potential market is just 1 sale and it’s not even clear that there’s a diminishing impact on the actual rights holder for that guy.
I concur that fair use could be an effective defense against infringement claims; the section you are referring mentions distribution/copying specifically, but it refers to other sections which apply to adaptation.
Regardless, the law itself seems clear that the right to adaptation is exclusive independent of distribution.
The fair use arguments are interesting, considering:
- AI businesses are commercial enterprises
- Many works (perhaps including an author's entire published output) were digested in their entirety to create LLMs
- If anyone can generate their own version of GoT/SoI&F with an AI system, does it diminish the market for the original?
(On the other hand, it's too bad HBO didn't have one at the time.)
It seems extremely unlikely that a company could successfully argue people writing fanfics for just themselves with the aid of an AI is sufficient to cause any harm to a business. The scale is minuscule. It’s likely not a substitute either.
Seems unlikely. You can hire someone to write you stuff that violates copyright, like fanfiction. It is unlikely that any court would find that the purpose / character of that use is commercial in nature; nor that the effect of creating it would have any effect on the potential market for the original.
Even if someone is generating a new novel for themselves to read every day, it doesn’t seem likely to pass those tests so long as they are not distributing it.
Selling this work would be copyright infringement. There’s no ambiguity there.
Generating it would not.