The deal to settle the Writers Guild of America strike last week could achieve more than to bring an end to one of the longest work stoppages in Hollywood history. It is likely to prove a benchmark agreement for other unions and professions looking to protect their members from the threat to their livelihoods from generative AI, not least the still-on strike Screen Actors Guild.
The WGA managed to wrangle firm limits on the studios’ ability to use of AI to generate scripts without the credited involvement of a human union writer while preserving enough flexibility, for themselves and the studios, to use the technology to increase productivity. And, as Cornell University professor of labor and industrial relations Adam Seth Litwin suggested in an op-ed in Friday’s New York Times, the balance the deal strikes could prove a model for workers beyond Hollywood looking to negotiate their own technological transitions, such as the striking United Auto Workers currently grappling with a transition to electric vehicle production.
Another party that could probably benefit from paying attention, however, is the U.S. Copyright Office, because the deal contains a potential pitfall for which the USCO is partly responsible.
According to the Memorandum of Agreement released by the WGA, writers will be able to use generative AI (GAI) as part of their preparation of a script so long as the obtain the consent of the studio that hired them. But, “The Company retains the right to reject the use of GAI, including the right to reject a use of GAI that could adversely affect the copyrightability or exploitation of the work.”
The purpose of that language is clear. Studios typically retain the copyright in scripts they commission or acquire, but they know that works created entirely by AI cannot be copyrighted in the U.S. Under the policy guidance the Copyright Office issued in March on the registration of works involving AI, however, works that contain a mixture of AI- and human-produced elements, could potentially be eligible for copyright depending on… well, it’s not clear.
According to the guidance, the determination of whether a work containing both human and AI elements is eligible for copyright protection will be made after the fact, on a case-by-case basis, based on how AI is used.
Here’s the relevant paragraph from the USCO’s March policy statement:
In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of “mechanical reproduction” or instead of an author’s “own original mental conception, to which [the author] gave visible form.” The answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work. This is necessarily a case-by-case inquiry.
Taken literally, that means neither the writer nor the studio can know for certain in advance whether the use of AI “could adversely affect the copyrightability or exploitation of the work.” Every AI-assisted script, in effect, will have been written “on spec,” with no assurance that it can or will be filmed.
Given the precedent of the “Zarya of the Dawn” case, in fact, is even possible that registrations once granted could be rescinded in whole or in part if the USCO later changes its mind.
Even if such adverse rulings prove rare, the ambiguity of the guidance introduces a degree of uncertainty into the preparation of scripts that benefits no one involved, except perhaps the lawyers, who will certainly find work litigating the inevitable disputes between writers and studios and challenging adverse rulings by the USCO.
The same will be true for the music, publishing and other copyright-based industries.
While the Copyright Office has always had the authority to reject applications based on a lack of sufficient creativity or originality, as well as for works produced by monkeys or other non-human authors, that authority has a clear statutory basis supported by more than a century of case law. The sort of line-by-line parsing implied by the guidance lacks that clear frame of reference, leaving it to creators and creative industries to guess where the line is.
That’s not a very good model for smoothly navigating a difficult technology transition.
In its recent Notice of Inquiry on AI and copyright, the USCO hints that updated version of its March guidance could be in the works, based on the skeptical feedback it received during a series of listening sessions it held over the summer. It could use a second look.