Which AI Components Are Copyright Protectable and Which Are Not?

Background 

On February 14, 2022, the Review Board of the U.S. Copyright Office denied a second request for reconsideration regarding a refusal to register artwork created by AI. Importantly, the application for registration indicated that the artwork was created "autonomously" by "a computer algorithm running on a machine." The applicant did not assert that the work was created with any contribution from a human author.

The decision addresses and denies the question of whether AI can be an author for copyright purposes. The decision, however, does not touch the question of whether and, in particular, what degree of involvement of AI in the creation process renders the output unprotectable under copyright law.

The Review Board's decision was not all that surprising. In the United States, the Copyright Act protects "original works of authorship" that are fixed in a tangible medium of expression. 17 U.S.C. § 102(a). Case law and commentary analyzing the authorship question in non-AI contexts have long suggested that a human is required for copyright protection. See e.g., Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 57-59 (1884) (holding that an author is "he to whom anything owes its origin; originator; maker; one who completes a work of science or literature"; describing a copyright as "the exclusive right of a man to the production of his own genius or intellect"); Urantia Found. v. Kristen Maaherra, 114 F.3d 955, 957-59 (9th Cir. 1997) (holding that a book containing the words "'authored' by non-human spiritual beings" can only gain copyright protection if there is "human selection and arrangement of the revelations"); Naruto v. Slater, 888 F.3d 418, 426 (9th Cir. 2018) (affirming dismissal of copyright claims brought by a monkey over selfies he took on a photographer's unattended camera; noting that the Copyright Act refers to an author's "children," "widow," "grandchildren," and "widower,"—terms that "all imply humanity and necessarily exclude animals").

The Situation in France, Germany, and the UK

The situation is similar in France, Germany, and in the UK where courts have never recognized any person other than a natural person as author of a copyrighted work.

The finding that AI cannot be an author for copyright purposes does not mean that AI-assisted outputs are void of any copyright protection. Works created by a human using software on a computer (e.g., Microsoft Word or Adobe Photoshop) are arguably protectable under copyright law. In its decision, the Review Board was careful to note that "the Board does not need to determine under what circumstances human involvement in the creation of machine-generated works would meet the statutory criteria for copyright protection." Feb. 14 Decision, at 3 n.3. 

English law provides that fully computer-generated works may enjoy copyright protection. It provides for a specific category for such works created "in circumstances such that there is no human author of the work" and defines the author as the "person" who has made the arrangements necessary for its creation. Such works are afforded 50 years of protection. The provision does not, however, directly address how a computer generated work could satisfy the classic requirement that copyrighted works be "original." English courts measure originality by reference to characteristics typically associated with human intellect—namely skill, labour, and judgment. The UK Intellectual Property Office recently closed a consultation looking into the provision, asking respondents how AI-generated works should be protected and whether they should remain protected in their current form, under a different scheme, or at all. The results are yet to be published. 

French and German courts also have yet to rule on what degree of human involvement might be sufficient to render AI-generated outputs protectable under the respective national copyright law. The more sophisticated an AI is the less likely it is that its outputs will be considered as work protected under copyright laws. In 2020, the European Commission published the final report on "Trends and Developments in Artificial Intelligence—Challenges to the Intellectual Property Rights Framework," confirming this finding at least for alpha numeric outputs (text). Other outputs, such as audio data and audio-visual outputs, may enjoy protection under related rights, and in the case of database protection, the sui generis rights might be available. Such protection does not hinge on originality. Of course, the software powering the AI and data used to train the AI also may enjoy copyright protection. 

Chinese Courts and "Human Intellect Imprints"

In China, although it is not clearly provided in current copyright law if AI-generated works are copyrightable, the Chinese courts have taken some initial positions by noting differences of AI-generated works with or without human involvement, and in some decisions recognized that AI-generated aspects with human intellect imprints could be protectable under certain conditions. For instance, the Nanshan District People's Court in Shenzhen ruled in what was featured as "China's first AI case"—Tencent v. Yingxun (2020)—that an article generated by Tencent's team by using "Dreamwriter" software has a certain degree of originality, and the process of the generation shows the "intelligent creation" of Tencent's team, and as such should be protected under the Chinese Copyright Law. 

This case certainly sends signals to incentivize the AI industry from a policy perspective. On the other hand, for AI-generated aspects without human involvement, e.g., any works generated by algorithms automatically evolved by AI through deep or machine learning, it remains to be seen how the copyright law will be developed to address copyrightability of such works. At least at the current stage, some Chinese courts continue to emphasize that creation and completion by natural persons should still be a prerequisite for works to be copyrightable.