Introduction
On August 30, the United States Copyright Office
published in its Federal Register a notice of inquiry requesting public opinion
on a variety of problems relating to the junction of copyright legislation and
computational intelligence (AI). The recent notice is the Office's latest step
in response to the plethora of copyright issues that have arisen in connection
with the use of artificial intelligence (AI) platforms, such as infringement
liability when instructing AI systems on copyright content and human authorship
obligations.
Increasing Friction between Copyright and Generative
AI Leads to Copyright Office Initiative
The Copyright Office's activities on AI concerns
accelerated in March, when the bureau announced the creation of a copyright and
AI program to investigate policy challenges connected to generative artificial
intelligence platforms. The declaration comes after the Office received two
registration applications last year, resulting in the agency's first attempts
to deal with AI-generated content. In the instance of Zarya of the Dawn, a
graphic novel using AI-generated artwork, the Patent and Trademark Office
eventually annulled the artificial intelligence-generated artwork portions of
author Kristina Kashtanova's copyright registration. The United States District
Court for the District of Columbia upheld the Copyright Office's refusal to
register the AI-generated image "A Recent Entrance to Paradise" owing
to the lack of a human creator in August.
Over the last year, copyright challenges raised by
generative artificial intelligence (AI) systems have dominated news headlines
in trademark circles. Several U.S. congressional investigations for the
relevant IP Subcommittees in the Senate and House of Representatives have
focused on generative AI. During those hearings, participants discussed the
impact of the federal right of publicity on artist protection, as well as
whether or not the use of content protected by copyright to train AI
constituted as fair use. Major generative AI firms, like as OpenAI, are facing
an increasing number of litigation alleging infringement of copyright and
unethical commercial practices, such as stealing user data. Proposed standards
for generative AI platforms in the European Union would demand the disclosure
of copyrighted content used for training AI models, potentially.
The Copyright Office is collecting opinions on four
policy areas based on previous engagements with agency participants on AI
problems, including those obtained through public listening events and
instructional webinars on the subject.
"(1) using copylefted works to train
Intelligence models;
(2) the legality of material generated by AI
systems;
(3) the possibility of accountability for infringing
works produced by AI systems; and
(4) the consideration of generative AI outputs that
replicate human artists' style or identity."
Agency’s Notice Explores Right of Publicity, Infringing AI Outputs and
Fair Use Issues
Concerning the use of copyrighted material by AI
systems, the Office acknowledged disputes concerning potential infringement responsibility
when such content is used to train AI. Representatives from Simplicity AI,
which claims that its use of freely accessible content is fair use, and Adobe,
which limited its training data to openly accessible and licensed photos,
testified at a Senate IP Committee hearing on copyright and AI in July. In
addition to feedback on infringement responsibility over training dataset
sources, the Office seeks advice on feasible recompense methods for artists
who's works are used in such datasets.
While the Office believes the law is unambiguous on
the human authorship requirements of U.S. copyright law, the agency is
investigating the proper breadth of copyright protection for AI-generated
works, including the amount of authority that a human needs to have over an AI
system for copyright to arise. Concerning infringement, the Office seeks
feedback on how culpability should be allocated when AI-generated works are
discovered to be substantially identical to the intellectual property used to
train the AI system.
The Office also invites public comments on state
right of exposure and unfair competition legislation, as well as international treaties
that are involved by AI systems that generate output that replicates a specific
artist's voice, likeness, or style. Legislative efforts to establish a national
right to publicity have not been universally supported by the artistic
community. Just this week, filmmaker and documentary groups expressed concerns
to the Senate IP Subcommittee that a right of publicity might limit works that
reference the name, likeness, or styles of famous individuals or well-known
artists.