“So, AI knows everything except that which the artist needs to know to determine whether their work has been used to benefit AI.” – Senator Peter Welch
The Senate Judiciary Committee’s Subcommittee on Intellectual Property held an oversight hearing yesterday in which Register of Copyrights Shira Perlmutter told the Subcommittee members that the Office is still working to get parts two and three of its promised report on Copyright and Artificial Intelligence out by the end of this year.
Part one of the report was published at the end of July and focused on the copyright implications of digital replicas created by AI. The report recommended passage of a federal law that would create a new form of property right for a person’s digital replica to disincentivize the creation of realistic but false depictions of individuals. On the same day the report was issued, a group of senators introduced a bill—the “Nurture Originals, Foster Art, and Keep Entertainment Safe Act of 2024” (NO FAKES Act)—to create a right for individuals to control digital replicas of their voice and likeness. Perlmutter said in yesterday’s hearing that she was pleased to see the bill and that the report and the bill are in agreement on all the most crucial aspects, such as that the right should cover all individuals, there should be limits on free transferability, and the creation of a safe harbor to incentivize the swift removal of unauthorized replicas.
The NO FAKES Act was introduced by Senators Chris Coons (D-DE), Marsha Blackburn (R-TN), Amy Klobuchar (D-MN), and Thom Tillis (R-NC) in July and would establish a federal property right for every individual in their own voice and likeness. There are exceptions for news, broadcast, advertising/ marketing and documentary use, public criticism or commentary and “fleeting or negligible” use, among others. An “actual knowledge” standard would be required to prove liability. The right is not assignable during the life of the right holder but is licensable for up to 10 years (or 5 years for minors under 18) and expires 70 years after the death of the individual. There are provisions for post-mortem transfer of the right and renewal procedures as well.
The bill would also establish a notice and takedown process for removing unauthorized replicas and would not hold platforms liable for linking or referring users to unauthorized content, “as long as the online service removes or disables access to the unauthorized digital replica as soon as technically and practically feasible after receiving notice of a claimed violation.” Online service providers must also designate an agent as the contact point for reporting violations.
Asked by Subcommittee Chair Chris Coons (D-DE) what keeps her up at night when it comes to the AI issue, Perlmutter said “the speed at which this is all developing.” In September during IPWatchdog LIVE 2024, Perlmutter told LIVE attendees that while she’s confident the issues around copyright and AI will eventually be solved, she’s “less comfortable about what it means for humankind.”
Perlmutter recently came under fire from Committee on House Administration Chairman Bryan Steil (R-WI), who sent a letter On Tuesday, October 29, to the Office asking for an update on the AI report, which Steil charged is no longer on track to be published by its stated target dates. Steil’s letter asked the Office to explain the delay in issuance of parts two and three, which Register of Copyrights Shira Perlmutter indicated in an oversight hearing by the Committee on House Administration would be published before the end of the summer and in the fall, respectively. “The importance of these reports cannot be overstated,” Steil wrote, explaining that copyright owners are relying on the Office to provide clear guidance. “The absence of these reports creates uncertainty for industries that are already grappling with AI-related challenges and hinders lawmakers’ ability to craft effective policy,” the letter added.
Perlmutter commented in the hearing that “we’ve been trying to set and follow our own ambitious deadlines” and the goal remains to get the rest of the report out by the end of the year, but that her key concern is to be “accurate and thoughtful.”
The forthcoming reports will include recommendations on how to deal with copyrightability of materials created using GAI and the legal implications of training on copyrighted works. The latter is most controversial and may in fact require additional legislation focusing on transparency requirements.
The Office received more than 10,000 comments on its Notice of Inquiry issued in August 2023 seeking public comment on the intersection of copyright law and artificial intelligence (AI). Notably, the comments were mostly original and substantive, rather than duplicates, making the review process lengthy. Perlmutter told the Subcommittee that nearly all commenters said the existing fair use framework is adequate to address issues surrounding the use of copyrighted works in training generative AI (GAI) models but that “views diverge strongly on what types of uses qualify as fair use.” There are presently approximately three dozen court cases pending that address this issue but no decisions. Once decisions start coming out, congress may be asked to step in, Perlmutter said.
In an exchange with Senator Peter Welch (D-VT), Perlmutter said the pushback from GAI companies on being transparent about what works have been used to train their AI models centers around what kind of transparency—how much and to what degree of detail. Some companies argue it would be impossible to identify all the works used.
“So, AI knows everything except that which the artist needs to know to determine whether their work has been used to benefit AI,” Welch commented. Perlmutter agreed that it’s important to have that level of transparency so copyright owners can make decisions about what they want to do.
Join the Discussion
No comments yet.