Is prompting with copyrighted article actually a copyright infringement?

Hey, I’m working on a new kind of service using OpenAI and I’m having a concern regarding the legality of commercially using generated content which is going to be prompted using third party’s articles.

Let’s say I want my trained GPT to summarise an article or a bunch of articles. Is it legal to use the outcome commercially?
Of course, the source and authors are going to be credited.

My intuition says that such use is allowed under the “News reporting” fair use exception well described on the copyrightuser website (I cannot post links, sorry)

What is your view on that?

It’s a big “it depends,” followed by a quick “if you have real concerns, talk to your lawyer.”

Generally, summaries of copyrighted works are not infringements. But if it is an extensive summary, more of an abridgement or condensation, then it is much more likely to be considered a derivative work and an infringement.

But, I’m not a lawyer and—more importantly—I’m not your lawyer.


Another piece of advice is to not add the copyrighted material to your system prompt or the knowledge files.
Otherwise making sure the input has been properly transformed before being outputted should help to avoid issues.

1 Like