Much ink has been spilled over whether copyright liability arises out of generative AI, both with respect to its training functionalities as well as any arguably derivative nature of its output.  Of potentially greater practical import to copyright litigants, however, is a question that has thus far garnered significantly less attention: can AI-generated works be copyrightable, and to what extent?

This open question of law, the resolution of which is likely to impact copyright litigation for the foreseeable future, is one with which both the Copyright Office and courts are currently grappling, with trends in policy actively emerging.

Copyright Office Policy and the Human Authorship Requirement

On March 16, 2023, the Copyright Office issued a statement of policy applying a human authorship requirement for copyrightability in setting forth a case-by-case approach as to whether AI-generated content is eligible for copyright protection.  According to the Copyright Office, that requirement derives from an inability of non-humans to qualify as “authors” under the Copyright Act, which protects only material that is the result of human creativity.

Applying this policy to the case of generative-AI and where AI technology creates works (whether written, visual or musical) in response to human prompting, the Copyright Office has taken the position that “the ‘traditional elements of authorship’ are determined and executed by the technology—not the human user.”  The Copyright Office thus concludes that such material is not the product of human authorship, and is not protectible under copyright.

Where, however, an artist materially modifies AI-generated content with alterations which otherwise meet the standard for copyrightability, or applies creativity in the selection or arrangement of AI-generated material sufficient to render the resultant work as a whole an original work of authorship, the human authored aspects of such works –  but not the aspects generated by AI – may be protected by copyright.

To effect execution of this policy, the Copyright Office requires that AI-generated material be disclosed when applying for copyright registration, and that previously filed applications which do not disclose the use of AI be corrected.  It also goes further, requiring the filing of supplemental registrations identifying and disclaiming any AI-generated material contained within previously-registered works.

Implementation of and District Court Challenge To Copyright Office Policy

The Copyright Office implemented its human authorship approach in the case of Zarya of the Dawn: a graphic novel for which generative-AI was utilized to create the comic book’s images.  There, and in evaluating an application for copyright registration of that work, the Copyright Office approved protection for the text and arrangement of images, but denied protection for the individual images themselves.  

In another instance involving AI-generated artwork titled “A Recent Entrance to Paradise” for which AI was named as the author and the developer of that AI named as the owner of the work, the Copyright Office refused registration, including in the face of the applicant’s work-for-hire argument, based on lack of human authorship.  That refusal and the associated human authorship policy being advanced by the Copyright Office is currently being challenged in the US District Court for District of Columbia in Stephen Thaler v. Shira Perlmutter and The United States Copyright Office (1:22-cv-01564), where cross motions for summary judgment are currently pending.

Implications for Copyright Litigants

Whether or not the Copyright Office’s human authorship policy remains in place, including to the extent it is upheld by the courts, the copyrightability of works generated by AI is bound to have a significant impact on copyright litigants moving forward. 

To the extent the human authorship requirement is upheld, the involvement of AI in the creation of asserted works is likely to become a regular avenue of discovery in copyright litigation and basis for challenging the validity of the asserted registrations.

This is likely to be especially true in the case of foreign works, for which a presumption of ownership and validity does not attach, and which may be particularly vulnerable to challenge given the varying policies with respect to the registrability of AI-generated content emerging across the globe, including in Berne Convention countries. 

Even for US works, however, discovery into the involvement of AI is likely, as are attacks on underlying registrations which do not comply with the disclosure, disclaimer or supplemental registration requirements set forth by the Copyright Office’s latest policy statement.  Copyright owners are thus well-advised to ensure they are staying apprised of and compliant with the requirements in force.

In the event the human authorship requirement is discarded allowing the registration of AI-generated works, additional questions implicating copyright litigants are likely to arise: 

Who is that author and/or owner of works created by generative AI – is it the AI, its developer, the AI owner/licensee, the human who prompted creation of the work at issue, or perhaps even the owner/author of works on which the AI was trained?  What are the implications for the work-for-hire doctrine under these circumstances?  And what are the ramifications for where liability arising out of such works can lie?

While many questions remain, one thing is certain: as long as the copyrightability of AI-generated works remains in question, companies are well-advised to consider the potential un-registrability – and thus unenforceability – of AI-generated works as they contemplate using AI tools in the creation of works they may ultimately want to protect.

Authored by Lauren Cury.