VENICE, Calif. – In her recent piece for VICE about an artificial intelligence-assisted fake celebrity-porn video created by a Redditor who goes by the handle “deepfakes,” Samantha Cole raised a variety of issues and concerns stemming from the ability of people to make increasingly believable fake porn videos employing relatively easy to use software tools.
One of the issues Cole did not address, however, was the intellectual property implications of such creations.
For starters, who owns the copyright to the celeb fake? For that matter, when the fake makes use of an existing work like the Nubiles video used in the Gal Gadot fake addressed by Cole, can such a fake legally be made without the permission of the rights-holder to the original work? Getting even more granular with the question, are there issues with media outlets using the modified celebrity images in their coverage of works created by the likes of deepfakes?
“Typically, the human operating the machine running the AI would be deemed the copyright holder,” attorney Larry Walters told YNOT. “But when the AI is modifying an existing, copyrighted work, the issues become more complex. The original work’s copyright owner has a right to control all derivative works. The AI operator is likely creating an unauthorized, infringing derivative work, in the example you raise with AI-assisted fake porn.”
I can already hear fans of an expansive reading of the exceptions to copyright screaming: “What about the fair use of parody?”
“There may be a fair use defense for some AI-assisted fakes,” Walters said, “but a full-length video changing nothing but the face appears to be an infringement. I could see some parody material staying within the realm of fair use, but the line is blurry.”
Copyright infringement isn’t the only potential pitfall of fake celebrity porn, of course. There’s also the question of the “Right of Publicity,” an intellectual property right with origins which are separate from the more commonly discussed (and litigated) realms of copyright and trademark.
“With the publicity rights claim by the performer, there is a recognized First Amendment-based, newsworthiness defense to those claims,” Walters said, adding that while the newsworthiness defense might help VICE should Gadot take exception to its use of screengrabs from the video, it might not shield deepfakes in the same way.
“The site reporting on the phenomena and using a short clip is likely protected, while the producer of the video may not be,” Walters said.
As technology improves and advances and AI enables people to make virtual porn which looks increasingly like the real thing, celebrity fakes will likely not depend on a copyrighted original work to achieve a great degree of verisimilitude.
While right of publicity claims may still be applicable to completely original works of fake celebrity porn (because the celebrity in question will still be able to argue the creator has used their “likeness”), the legal issues will only get muddier when creators obtain the ability to make highly realistic virtual porn from scratch.
Of course, the same can be said about a myriad of sociocultural issues presented by highly realistic fakes, as well. Just imagine the chaos (and extortive possibilities) which could ensue if someone were to create some highly convincing fake security camera footage of a famous person committing a heinous crime, for example.
The simulated activity depicted needn’t be criminal to be incredibly damaging, of course. No video of a politician you hate being urinated on by hookers exists to impugn his or her character? No problem; just create one yourself with the help of AI and existing footage of a golden shower group session!
“The IP issues associated with AI have fascinated lawyers like myself for some time now,” Walters said. “Courts will struggle with this issue for years to come. Ultimately, people will start to question whether anything is real, as AI gets more powerful. That has broad implications for law, politics, and society.”
