James Grimmelmann explains how AI is changing copyright

So when you write a novel, when, say, J.K. Rowling writes Harry Potter, she gets a copyright for the novel, and she also gets the copyright-- that, that copyright includes not only the right to make, uh, exact copies of Harry Potter but also derivative works.

If somebody else wants to make a novel set in the Harry Potter universe, they need permission from J.K. Rowling to do that.

Something similar works with software.

You know, when Microsoft creates Microsoft Word, they get a copyright on Microsoft Word.

I can't make a copy of it.

I also can't take Microsoft Word and modify it in some way.

Even though I modify it a lot, add a lot of features, remove a lot of features, I can't publish that without Microsoft permission.

So this is the kind of economics underpinning of the, much of the software industry.

You know, some of these, these copyrights are worth billions of dollars.

But a tricky thing is that the, what counts as a derivative work can be a little bit complicated because, um, say, we probably don't want Microsoft to have a monopoly over the ability to read and, and produce doc files.

And so figuring out what kind of, uh, similar software can you make, what it counts as a derivative work is tricky.

The last, uh, year or two, we've seen the AI models really throw a lot of this into chaos.

So when you write a novel, when, say, J.K. Rowling writes Harry Potter, she gets a copyright for the novel, and she also gets the copyright-- that, that copyright includes not only the right to make, uh, exact copies of Harry Potter but also derivative works.

If somebody else wants to make a novel set in the Harry Potter universe, they need permission from J.K. Rowling to do that.

Something similar works with software.

You know, when Microsoft creates Microsoft Word, they get a copyright on Microsoft Word.

I can't make a copy of it.

I also can't take Microsoft Word and modify it in some way.

Even though I modify it a lot, add a lot of features, remove a lot of features, I can't publish that without Microsoft permission.

So this is the kind of economics underpinning of the, much of the software industry.

You know, some of these, these copyrights are worth billions of dollars.

But a tricky thing is that the, what counts as a derivative work can be a little bit complicated because, um, say, we probably don't want Microsoft to have a monopoly over the ability to read and, and produce doc files.

And so figuring out what kind of, uh, similar software can you make, what it counts as a derivative work is tricky.

The last, uh, year or two, we've seen the AI models really throw a lot of this into chaos.