This episode is the final one of the season in collaboration with the Engelberg Center on Innovation Law and Policy at NYU Law. This season we’ve talked to lawyers, artists and other scholars who have helped us unpack some of the thorny issues for those working in art and code as they unleash their work into the world.
When we first began our research for the season we started by speaking to Michael on a range of legal issues pertaining to creators and artists working today. Then I asked what I thought was a straightforward question and was amazed at the response. I asked Michael about generative AI, and especially things like artwork created through machine learning algorithms. It was at that point that I first learned of the Monkey Selfie Copyright Dispute.
In our previous episode Michael explained to me that the courts had ruled that works that are not created by a human cannot be copyrighted. And furthermore, that the U.S. Copyright Office had advised that this legal case would impact work created with generative A.I. When I first learned this, I was surprised and intrigued. Since that previous interview, Michael got back in touch to let me know that we finally had further legal cases relating to generative A.I. that specifically cite the monkey legal case as precedent. We revisit our earlier conversation then catch up with Michael for an update on these recent cases.
Michael Weinberg is the Executive Director of the Engelberg Center on Innovation Law and Policy at the NYU School of Law. His research centers on open source, open access, and innovation. He is also the Co-Director of the glam-e lab, a project that uses direct representation to develop model policies and terms for cultural institutions that are creating open access programs.
Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence
US Copyright Office on Zarya of the Dawn
This season of the podcast was produced with the Engelberg Center for Innovation Law and Policy at NYU.
Our host is Lee Tusman. Our audio production is by Max Ludlow.
All of the music on today’s episode are by our audio engineer Max Ludlow. The tracks are Body Memory, Poole and Relic, CC BY.
This episode is licensed under CC BY 4.0
Lee Tusman
You’re listening to Artists and Hackers, a podcast on art, code and
community. We talk to programmers, artists, poets, musicians, botmakers,
students and now legal scholars in an effort to look at online artmaking
and a history of technology and the internet. We’re interested in where
we’ve been and speculative ideas on the future.
Today’s episode is with Michael Weinberg, the Director of the Engelberg Center on Innovation Law and Policy at NYU. In our earlier episode, we covered a range of legal issues pertaining to creators and artists working today. Then I asked what I thought was a straightforward question and was amazed at the response. I asked Michael about generative AI, and especially things like artwork created through machine learning algorithms. First we’ll listen to our earlier conversation on the topic, then I’ll come back and speak with Michael again, for an update on the latest legal issues in generative art.
So one of the strange things that’s happening is artists and software writers, programmers and other people are making tools that are used to make artwork. Let’s say I make a tool, maybe I make some new kind of digital painting software, or I write a machine learning library that others can use to generate poetry or plays or dances. What are some of the structures around what I can kind of specify in terms of who uses that software to make artwork. If I make this tool to make digital paintings, do I own the digital paintings that other artists that use my software make? Is there any restrictions on the work that they make or that I can even say about what they make with my software?
Michael Weinberg
First off, owning the copyright in a tool does not necessarily or
automatically give you ownership in the things that are made with the
tool. If I type something in Microsoft Word, if I write a book in
Microsoft word, Microsoft does not own my book. Now we’re using podcast
recording software to record this interview. The company that makes the
software to record the interview does not own a copyright in the
podcast. Or the company that makes the editing software you use to edit
everything together and and remove all of my ums and stammers does not
own the output, the thing that you create with the tools. Now that
distinction can feel like it breaks down a little bit when you have some
of these, especially some of these machine learning-based tools that
feel like instead of a person using the tool to create a work it’s kind
of the tool doing all of the work itself and there isn’t really a person
involved. There’s a little bit of debate, I mean there’s a little bit of
debate in legal world as to what should happen in that case. I think the
best answer and the answer that has been the case up until now and
probably will continue to be the case for the foreseeable future is that
you need a person to have a copyright. And so if it is actually true
that just a robot made whatever the work is, that’s great. The work
exists in the world but the robot doesn’t own the copyright. And this is
actually very similar to a case that was high profile or at least high
profile in copyright legal nerd world a couple years ago where it was
called the “monkey selfie case” and it was a situation where there was a
picture that went viral on the internet. And the story of the picture
was that some monkeys had stolen a photographer’s camera out of his bag
and taken a bunch of pictures of themselves because they were… or they’d
come up to a camera because they were seeing their reflection in the
lens and they were kind of fascinated with it and so they were they were
taking pictures of themselves basically. And so it went viral. The
pictures were very cute. And then the photographer tried to assert
copyright over the pictures. And the court, and ultimately this is the
copyright office, has put this in their big book of copyright rules,
it’s called the Compendium of Copyright. The monkeys took the picture,
but monkeys aren’t people and so monkeys can’t have copyright. And so it
feels like monkeys taking selfies is very different from an elaborate AI
machine learning algorithm generating art. But I think the principle was
the same which is you need a person involved in order to be able to have
a copyright in the first place. And there are lots of good reasons for
that. One of which is if you want to use that photograph or the image or
whatever it was generated, you need a license from somebody. And you
know a monkey can’t give you a license, can’t enter into a licensing
agreement. A robot can’t enter into a licensing agreement. So there just
needs to be a person for the system to work at all.
Lee
What about the person that wrote the robot software or… I guess I’m
trying to push at the limits here. I’m trying to figure out where does
that end and the robot begins.
Michael
Yeah, right? I mean this is a good question. And I think it becomes a
little bit of a fact specific question. If the person who wrote the
software is really the driving creative force behind the output then you
can probably argue that the person who wrote the software is sort of
working all the way through the process. If you can really draw a
straight line between the person who wrote the software and the output
then maybe there’s an argument for that. But in that case, the the
software isn’t really a tool in the sense that you know a word processor
is a tool, right? It’s more kind of part of the work that they are
doing, and it’s a tool being manipulated by the creator as as she’s
creating it. potentially if the creator, if someone creates a tool and
then someone else comes along and uses the tool to achieve some sort of
creative output, in that case, it probably is the case that that person
who’s using the tool is providing the creative spark and so maybe
they’re the person who should own the copyright and you can make up
hypotheticals that really kind of finally, that draw really fine
distinctions here and those hypotheticals will probably pop up in
reality over the next couple years. But I think you still get to a
situation where the fundamental question is, who is the person who is
ultimately responsible for the creative spark behind this specific
work?
Lee
And that’s where we left off with monkey selfies and generative art in
episode 20 when I spoke to Michael last year. But recently he let me
know that we finally have legal cases playing out in courts pertaining
to the monkey selfie case. Ever since I’d last spoken to him I had
brought up the story to family and at parties, I’m pretty sure my
friends are sick of hearing about it. But now, the monkey selfie case is
back in the news, at least legal news, just as predicted. So I spoke
again with Michael, to find out the latest.
Michael
So I think the good news is that everything that we talked about earlier
is still correct and everything we talked about is still right, and
these these questions that we were talking about of the line between
when the human is the creator and when the monkey or the AI or the
machine is the creator are still really important questions and really
key questions. And the thing that has changed since we last talked is
that we’ve actually seen in the US, the US Copyright office weigh in on
some attempts to register copyrights for works that were created either
by AI or in collaboration with AI and so it gives us a little bit more
context. It makes the hypotheticals a little bit more concrete to
understand how the lines are drawn.
Lee
And what are some examples of these?
Michael
So there are two high profile examples. One is an example where someone
tried to register a picture that they said was made by their AI program,
a generative AI program. And so they wanted to register the picture, and
the author they wanted to be the generative AI program itself. So not
someone who said I made this by prompting generative AI and therefore I
want to be to owner of the copyright, but someone who said I think the
robot is the author 100% and the copyright office said no. Robots cannot
be the authors. AI cannot be authors. You need a person in the process,
touching on that monkey selfie case that we talked about before. And so
it’s clear right now that the copyright of the US copyright office
position is that AI, and you know robots or software alone cannot be the
creator of a copyright protected work. It can create paintings but those
paintings, or anything else, don’t get copyright protection. They’re
just right now, they’re just in the public domain.
Lee
And can you say a little bit more about that? Did they explicitly state
that that’s the case or just because it’s not possible to copyright them
we or lawyers understand that to mean that it’s in the public
domain?
Michael
Yeah, I mean they rejected the registration. They said it’s not eligible
for copyright protection. This is now being litigated. The person who
submitted the work, the painting, is someone who is I think it’s safe to
call him a kind of an activist who has tried to have his computer
programs get patents in the past. he’s someone who really wants his
software to have the full scope of creative ability within the law and
so the copyright office was very explicit. They rejected the
registration. They said that there is no, there’s no human creator and
so this is not eligible for copyright protection. The person who tried
to do it sued and appealed that and it’s now kicking around the
appellate court. But it’s been very explicit, if we’re just talking
about generative AI there’s no human and so there’s no human to own the
copyright. No one owns the copyright in the work.
Lee
Does that have implications for how people use things like Dalle?
Michael
I think it does. I think they’re really potentially very interesting
because what it suggests pretty explicitly is that the works coming out
of Dalle are not eligible for copyright protection, the works coming out
of Dalle are in the public domain. And so right now we have this
situation where we haven’t had for like a century where brand new works
are entering the public domain directly. They’re not being protected by
copyright for any period of time. They’re just generally available to
anyone to do whatever they want with.
Lee
That’s incredible! And I’m curious what the second legal case is that
you had mentioned?
Michael
So the second legal case was a little bit more nuanced. This was someone
who had made a graphic novel and they had written the novel themselves
but they had basically turned to Dalle or one of the other generative
AIs to create the images in the graphic novel. And they had registered
the graphic novel itself and there was a little bit of back and forth
but eventually the copyright office rejected copyright registrations for
the images. So the copyright office said you the person did a lot of
work. The generative AI was also very involved, but the generative AI
doesn’t get to be a co-author from a copyright standpoint. But you as
the person who wrote the book, you own the book. And also if you think
of the book, a graphic novel as a compilation, a thing that combines the
text and those images. Those images are positioned on the page a
specific way. They’re kind of ordered in a specific way. That entire
work, you can be the sole author of, but you can’t claim copyright in
the images themselves. And the AI doesn’t get copyright in the images
themselves either. And so if someone else wants to come along and copy
the entire book they’ll be infringing on the author’s copyright. But if
you take the images alone out of that graphic novel, those are not
protected by copyright. Those are in the public domain. So this case is
a little bit more of an example of that kind of collaborative approach
that we had touched on. But it’s still, you get in the same place,
right? The parts of the creativity that the person was directly involved
with, those are eligible for copyright protection. The parts where it’s
mostly generative AI, those are not eligible for copyright protection.
And I think what we’re going to see over the next couple months and
years is where that line is between, you know the human parts and the
generative AI parts.
Lee
Well that’s fascinating. Michael thank you so much for updating us. I
appreciate it.
Michael
Thank you so much.
Lee
This was our final episode of the season, which was made possible by
support from the Engelberg Center on Innovation Law and Policy at NYU.
Prior to this season, I had a more limited knowledge of the legal issues
we covered this season, despite working in new media art and music for a
long time, and I thought that other folks would want to learn more as
well. While we’ve only scratched the surface, we hope to cover more of
these areas going forward, not only because new ethical, social and
legal issues constantly come up when working with new technology, but
also because they of course suffuse our society overall.
I was genuinely surprised by the monkey selfie legal case, especially when Michael first presented the idea, that a guidance relating to a monkey that cannot legally hold a copyright could be the key to interpreting legal decisions based on algorithms. So I was surprised again when Michael said this decision had been upheld and the implication was that new generative works created through these kinds of machine learning algorithms are new works, are considered to be in the public domain. Especially since we covered the public domain in episode 22 where we spoke with Brewster Kahle from the Internet Archive where he spoke about the law of copyright being automatically applied to almost all works, for 70 years after a creator’s death, unless an author explicitly waives or grants exemptions.
In my first interview with Michael, he ended our conversation with an idea that has stuck with me. He said, “when you’re making decisions, try and empower the good users and not spend a lot of time trying to limit the bad faith users. Think about who you want to be using this and how you want them to be using it and put in place the licensing structure that allows them to do it.”
When speaking with Dr. Jane Anderson and with Courtney Papunee in episode 23 I was inpsired by the innovative creation and testing by Local Contexts, which grows out of needs of indigineous and local communities where they’ve been ignored or mistreated by unsupportive laws. It’s an example of how communities are working to ensure their culture, community, heritage, even their data are used or presented respectfully.
Local Contexts, and initiatives like Creative Commons, free and open source and ethical open source licenses, are genuinely exciting, important tools, that I hope will strengthen artists and creators work and help lead to the creation of healthy and supported creative communities.
If you have a response to this episode or the season as a creative practitioner, researcher or scholar, or are working on new initiatives, licenses and tools, please get in touch to let us know about it.
This season of the podcast was produced with the Engelberg Center at NYU Law. My name is Lee Tusman and our audio producer is Max Ludlow. Our designer is Caleb Stone.
All of the music on today’s episode is by our audio engineer Max Ludlow with the tracks Body Memory, Poole and Relic.
You can find full episodes, transcripts and links to find out about our guests and topics on our website artistsandhackers.org. You can find us on mastdon at artistsandhackers at post.lurk.org and you can always write to us on our website. Please leave us feedback wherever you get your podcasts. Thanks for listening.
