2 Creative IP Attorneys On The Complications Of AI-Generated Art

With artificial intelligence taking on an increasingly prominent role in the creative industries, intellectual property attorneys who work with creative clients are monitoring the slowly developing body of legal guidance regarding AI and adapting their practices to match. 

Still, challenges abound. We recently spoke with Emily Poler, founding partner at New York City litigation firm Poler Legal, and Katerina Duarte, founding partner at the New York City art–and-media-focused litigation law offices of the same name, to better understand the complications that arise from dealing with AI-generated art and how creative professionals can protect their work.

How AI Could Change the IP Landscape

Poler says she anticipates changes to IP licenses to account for AI, such as clauses in license agreements detailing “who is responsible for policing any potential infringement including any infringement by an AI platform.”

“As far as litigation, I think we’re going to see not just infringement actions, but also actions for defamation when AI generated content just gets the facts wrong, as well as right of publicity cases stemming from photos or videos that incorporate a celebrity’s image,” she adds.

Duarte says that her smaller creative clients in particular feel an emotional punch after discovering AI-generated works based on their original creations.

“For the little client, it’s very personal. This is your character; this is your artwork; this is someone that you’ve known for five years as you’re developing your book or your comic strip, and then someone just takes it, right?” she says.

Still, there isn’t always much Duarte can do for her clients — especially given the international nature of some potential copyright violations.

“I’m not going to take a retainer to send a letter or try to sue [the alleged IP violator] when a lot of these people that are creating the software or creating the modified or derivative works are abroad and are judgment proof,” she says.

Separating the Artist From the Tools

One particular challenge for enforcing IP protections against AI-generated work lies in determining what party is responsible for the possible legal fallout resulting from art created by AI, Duarte says.

She raises an important question for those defending their IP against derivative AI works and for those using AI to create supposedly original works: If the user just drafts a prompt and clicks a button, did they really create the work? Was it, instead, the engineers who created the AI software, or perhaps the software itself?

Put more simply: Where does the artist’s creative input end and the software’s output begin?

AI users, Duarte suggests, could potentially take the position that AI software companies should take the blame if, for example, an image generated with their software is at the center of an IP dispute.

“You created a condition where you are allowing people to infringe on other people’s creative expression,” she says of the software companies that create the programs allowing for AI-generated art. “How could you not be liable for that?”

To take the question further, Duarte points to past legal contentions over IP protections for new technologies like photographs or digital art.

“This can go on forever, right?” she says. “If I’m using someone else’s software to create my art, but I’m licensing their software, do they have a claim to my artwork? What if they stop the license?”

Waiting on the Courts

Despite the risks their smaller creative clients currently face, Poler and Duarte both suggest that real anti-AI intellectual property enforcement won’t be possible until the federal courts step in with rulings specifically regarding AI.

“Also, at least in the U.S., the First Amendment and state laws governing the right of publicity will play a role in shaping the contours of the law governing the use of AI,” Poler adds. 

“The First Amendment will come into play when there are issues over whether a new work is transformative, newsworthy, or a parody, and the right of publicity will come up where AI is used to make a celebrity ‘appear.’”

The U.S. Copyright Office has issued at least some guidance regarding AI-generated art and copyright protection, stating that outcomes for individual cases “will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work.”

New developments may come sooner than expected thanks to several ongoing lawsuits. In January of this year, a group of visual artists alleging copyright infringement sued several AI companies whose products generate digital images.

In early July, comedian and author Sarah Silverman, as well as authors Christopher Golden and Richard Kadry, launched lawsuits against Meta and ChatGPT parent company OpenAI alleging that both organizations used datasets from so-called “shadow libraries” that illegally contained the authors’ protected works to train their AI products.

How Creatives Can Protect Their Work

Until the dust has settled, there are a few tactics creative professionals can use to protect their work. 

“First, watch out for others using your intellectual property rights (or name or likeness), and don’t sit by if you see unauthorized uses,” Poler suggests. 

“Second, talk to a lawyer! Really! I know it seems self-serving, but not infrequently there are ways for creators to save themselves a world of hurt by talking to an attorney before they launch a project or respond to someone else’s project.”

For creative professionals who aren’t sure about a particular legal issue, Duarte suggests reaching out to an organization like Volunteer Lawyers for the Arts.

She also adds that creative professionals who are concerned about AI users creating works based on their original work should be careful about which works and how much work they put on the internet.

“Guard your work, keep it tight, and think of other ways to sell it and make money from it. Don’t necessarily put it in a digital format because, if you don’t put it online in a digital format, it’s not going to be copied as readily,” she says.

Avoiding digital media may prove difficult for creatives operating in a digital world however, and there’s no guarantee that Silverman and Co.’s works wouldn’t have been included in the datasets allegedly used by Meta and OpenAI if they had only been published in physical formats.

For example, PDF scans and HTML versions created and uploaded by fans (or even university professors looking to help their students access works for academic purposes) to shadow libraries like Library Genesis, ZLibrary, and Sci-Hub can still expose the works to infringement risks.

---

Ethan Beberness is a Brooklyn-based writer covering legal tech, small law firms, and in-house counsel for Above the Law. His coverage of legal happenings and the legal services industry has appeared in Law360, Bushwick Daily, and elsewhere.