For a free audio version of this blog, click here.
It is probably not a bright idea for me to write this post. I finished it a few days ago, and things changed. You Patreon folk will see it at least a week before the website folk, and by then, I expect everything to be different again. So I’ll be updating it. Those of you hearing it on audio will hear a version recorded in the middle of September. Sorry about that.
At the beginning of September, Can Stock Photo announced it was going out of business. I have used Can Stock Photo since 2009 to source the art you usually find at the top of this blog. When AI art became good enough to pass as regular art, I got a lot more cautious with the downloads for reasons I’ll discuss later in this post.
What I did on Can Stock Photo, and apparently, what I’ll do going forward on a different site, is that I would only use art uploaded before 2020 or so. That way, I knew the art was the work and property of the artist, and not created by some website that more than likely stole someone’s artwork to train the bots.
I can’t improve on Can Stock’s words in their closing notice:
After nearly 20 years in business we have been forced to make this very difficult decision and would like to thank our many thousands of talented contributors and customers for making it possible. The industry has changed significantly over this time, with CanStock launched in 2004 during the early mass adoption of digital cameras, and before ‘social media’ was even a phrase. Today of course everyone has a capable digital camera in their pocket, and the advent of AI means amazing images can be created for free from programs with just a few keywords. Decreased business and increasing costs has made it no longer possible to keep operating, to our great disappointment.
In theory, the licenses that every person who used Can Stock Photo agreed to will remain in effect, although the prudent thing to do is download the license for each piece of art used, particularly on those used in for-profit ventures, like covers and the like.
You should be doing that anyway. When you contract with an artist, it should be a valid contract. When you download art from a commercial site, you should also download the license for that art. It’s easier and better for all around to keep these licenses in two files, one on your computer and one in paper…because, yes, folks, computers die, programs disappear, and sometimes you can’t access something you downloaded ten years ago because the programs have changed so much that the files are no longer compatible.
What remains compatible? Paper.
Why would you want to do this? Well, we’ve been reviewing our licenses at WMG Publishing, some twelve years old, because we’re opening stores with merchandise. Most of our covers come from artists like Philcold, whom we found on Dreamstime, and then worked with directly for projects like my Diving series and Dean’s Seeders series.
Not every license allows for merchandise. Many don’t even allow book covers over a certain number of sales. A lot of writers who are careless about such matters simply make an assumption that they have the right to use these things.
They don’t. And to make matters worse, the licenses and terms change all the time. So if you go on, say, Dreamstime, to look at the license today for something you licensed in 2015, today’s license does not cover the 2015 download. What covers the 2015 download is the 2015 license that existed on the day you downloaded the art.
Yes, some sites save their previous licenses, and some sites don’t. It is up to you to know what, exactly, you licensed.
Then you need to understand it. It isn’t something that’s a one and done. Copyright law shifts all the time. That’s why I took a two-semester class in copyright law at University of Nevada, Las Vegas, in 2021. I know copyright, but I don’tknow all of it. I have a current copy of The Copyright Handbook, and I use it. I also keep track of various court cases and read a lot of law blogs on copyright.
That does not make me an expert or an intellectual property attorney. I’m not even close. But I am informed because this is my business. It’s my job to stay ahead of everything that’s going on.
And a lot is happening on the copyright front with the copyright of works that incorporate artificial intelligence.
In an important case that came down in August, Judge Beryl A. Howell of the US District Court for the District of Columbia ruled with the copyright office and against a computer scientist who claimed that a two-dimensional piece of art created by his AI program deserved copyright protection.
Howell found that “courts have uniformly declined to recognize copyright in works created absent any human involvement,” citing cases where copyright protection was denied for celestial beings, a cultivated garden, and a monkey who took a selfie.
The ruling is the first in the country to examine whether or not AI art is copyrightable. That’s iimportant because, for now, that and the ruling from the copyright office are the guide. As Howell herself noted in the decision, there’s a lot more for the courts to decide. She wrote:
Undoubtedly, we are approaching new frontiers in copyright as artists put AI in their toolbox to be used in the generation of new visual and other artistic works. The increased attenuation of human creativity from the actual generation of the final work will prompt challenging questions regarding how much human input is necessary to qualify the user of an AI system as an“author” of a generated work, the scope of the protection obtained over the resultant image, how to assess the originality of AI-generated works where the systems may have been trained on unknown pre-existing works,how copyright might best be used to incentivize creative works involving AI, and more.
But these issues aren’t settled yet, and might not be settled for years. The way that novel questions of the law get settled in the United States is for them to wend their way through the courts. Sometimes the appellate courts agree, and then the issues are decided.
Occasionally, though, the issues are so complex and the courts are so divided that the issues will make their way to the United States Supreme Court. Unless that happens in an expedited manner, issues like that won’t see a resolution for eight to ten years, maybe more.
Copyright is particularly dicey, because it has—in the States—financial damages that are set by statute. Which means this: If an infraction is proven, then the court has very little leeway about how much the copyright violator must pay for their actions. The court can’t, say, charge the violator a dollar per infraction. Nope. That amount is set under law.
Right now, one of the unsettled questions is this: If existing art is used without permission to train AI, is the artist entitled to compensation for that use? Artists believe they are, and are suing for that. These mega corporations developing AI believe that they have fair use on their side.
The thought experiment that every writer who wants to use AI-generated art on their covers (and inside their special edition books) is this: If the courts decide for the artists, and the training is an infraction of copyright, who pays the fines? The megacorp, clearly. But who else? Anyone who downloaded a piece of AI art based on that illegally obtained training?
We don’t yet know.
That’s all too fraught for me.
Howell’s decision is a little more prosaic. It determined this: If I use AI art somewhere, then I can’t copyright that piece of art.
However, I might not be able to copyright the entire book. (Again, that needs to be litigated.) Everything might end up in the public domain—which means that anyone can use my book and that art without compensating me.
I find that really unacceptable. Writers—artists—should be paid for their creative endeavors. This is one of the many reasons I support the Screenwriters Guild and SAG-AFTRA in their strike against the studios. I wrote this post before the writers strike portion of this was settled. The WGA got something of what it wanted on AI, but not all. I’ll examine that in depth in a future post.
Other developments continue in the AI legal front. On September 19, seventeen writers, including George R.R. Martin, John Grisham, and Jodi Picoult, joined the proposed class action lawsuit against OpenAI. From Variety:
The lawsuit, filed Sept. 19 in the U.S. District Court for the Southern District of New York, seeks an injunction blocking OpenAI from continuing to use the authors’ works to train ChatGPT, as well as unspecific monetary damages (and statutory damages of up to $150,000 per infringed work).
This lawsuit is potentially huge. Anyone using ChatGPT for anything could be in the same kind of trouble we discussed above for the suits discussing the art.
For the legal geeks among you (like me), here’s a link to the complaint.
The legal action isn’t all in the U.S. The European Union is working its way toward finalizing a comprehensive legal framework for AI. Last I saw, the EU AI Act is in a three-way negotiation among its representative parts to finalize the act.
The act will cover everything from biometric surveillance to transparency of use to generative AI—which is the part that currently interests me.
That all sounds unimportant to most Americans who have no idea that what happens in other countries has an impact on them. As Alex Engler of the Brookings Institute pointed out on a panel at Stanford last summer, international corporations don’t want two sets of regulatory standards. He said,
Corporate interests will fight tooth and nail if you have two dramatically different standards for online platforms.
Which set of standards will prevail? That depends on a lot of things. Sometimes it is the most stringent, because if a company complies with that standard, they automatically comply with others. Sometimes the first to regulate is the one that will set the standards.
On this issue, it looks like the E.U. will be setting the standards. It will be the most stringent, and it will also be the first (at least that I can find) to establish widespread regulations.
But we don’t know. When things are up in the air, it’s better to avoid them. You want your work to be as legally protected as possible, so you need to err on the side of caution.
At least, that’s how I function.
We’re seeing a lot of fallout on this issue, though. Can Stock Photo is just one company that’s running into trouble. They’re having difficulty finding people to use their site, even though they seem to be clear on AI generated art.
I’m sure other places are having difficulty as well. I know some of my artist friends are. Some of them have had a real tough time trying to figure out if they want to spend the legal fees to join the existing lawsuits. I’ll never forget the anguish I heard from one of the best artists in the business when he realized his distinctive art was used to train AI. If you put his name into an AI generator, you will get art that is clearly composed of images stolen directly from his paintings.
So here we are, and we remain in flux. The D.C. decision is just the first. The OpenAI lawsuit is just getting started, and the EU’s actions are also just the beginning.
I’ll keep an eye on all of this, and report to you from time to time.
But let me urge you to act as I’ve urged all along.
Learn copyright law. If something is unsettled, avoid it.
Be cautious. It’s better to use a tool once that tool is judged safe to use. That way, your own art and writings will be protected—as will your bank account.
This weekly blog is reader supported.
If you feel like supporting the blog on an on-going basis, then please head to my Patreon page.
If you liked this post, and want to show your one-time appreciation, the place to do that is PayPal. If you go that route, please include your email address in the notes section, so I can say thank you.
Which I am going to say right now. Thank you!
Click paypal.me/kristinekathrynruschr4e to go to PayPal.
“Business Musings: AI Update: Copyright And Other Things,” copyright © 2023 by Kristine Kathryn Rusch. Image at the top of the blog copyright © Can Stock Photo / focalpoint.