The relationship between AI and patent, copyright, trademark, and fair use

Just about everyone has now heard the term "generative artificial intelligence (AI)." But how AI will be integrated into our existing legal framework is growing more complex, and more difficult for the courts and Congress to address.

Thus, at the close of last year President Biden issued an executive order directing federal agencies, including the USPTO in consultation with the Director of the Copyright Office, to issue recommendations and take action on AI policy in 2024.

On Feb. 13, 2024, in compliance with the President's directive to the USPTO Director to issue guidance to patent examiners on AI-assisted inventions, the USPTO published "Inventorship Guidance for AI-assisted Inventions." The guidance declared that "AI-assisted inventions are not categorically unpatentable for Improper Inventorship," so long as a human significantly contributed to the invention, even if AI "may have been instrumental in the creation of the claimed invention." The USPTO supported this guidance by relying on the current statutory framework, which "only require[s] the naming of the natural persons who invented or discovered the claim invention, irrespective of the contributions provided by an AI system."

The USPTO's comments appear to offer clear guidance on AI integration in patent law. Simply put, AI can assist an inventor, but a human must significantly contribute to the invention. Of course, the question may be what constitutes a "significant" contribution, but for now, the guidance has at least answered the question of whether AI-assisted inventions can be patentable, and they can.

The same guidance is urgently needed in copyright law where the predominant questions are whether 1) AI-generated content can be copyrightable - in other words, can AI replace traditional human authorship; and 2) whether AI's ingestion of data and authored works for learning is fair use. The law so far indicates that AI-generated content is not copyrightable. But no decisions have ruled on whether AI ingestion of data is fair use.

Output - When Does AI Overtake the Human Author?

In Thaler v. Perlmutter, on appeal to the U.S. Court of Appeals for the D.C. Circuit, the lower D.C. District Court found that the Copyright Office did not err in finding Dr. Stephen Thaler's artwork as ineligible for copyright protection because it was "autonomously created," meaning it was generated by an AI system without direct contribution by a traditional human author. Thaler developed his own generative AI system, the "Creativity Machine," which he then used to generate the artwork in question. In a brief to the appellate court last month, Thaler argued that a creative work generated by the creator and user of an AI system should be copyrightable. Citing to the statutory language and the Supreme Court's reading of it, Thaler argues the Copyright Act explicitly allows for non-human authors to accommodate corporations as authors. Thaler argues this allowance is a means of providing public access to copyrighted works. Despite this, Thaler laments that the Copyright Office relies on outdated cases, decided before we could fathom the potential capabilities of AI, rejecting registration for "non-human spiritual beings" and animals.

It remains to be seen whether the D.C. appellate court will affirm the lower Court's position on the copyrightability of AI-generated or assisted works, or whether they will take into account the USPTO's "Inventorship Guidance for AI Assisted Inventions." But, unlike the USPTO, the Copyright Office has not yet issued its recommendations to the President, leaving those utilizing AI to create these kinds of works and the artists and authors being imitated by AI to make unguided assumptions on what is and is not copyrightable and what is and is not infringing.

Input - Fair Use v. Collective Licensing

Next, there are fundamental disagreements on whether AI data scraping constitutes fair use. Some are of the opinion that AI's data scraping for learning is fair use because it is akin to human learning, where AI ingests material to gain an understanding of the underlying principles and facts that govern the world and outputs an original or transformed expression. Conversely, those who advocate for a licensing regime believe that AI learning is akin to copying and pasting existing text, creating an output that is like a collage made up of a million pieces of original expression and thus copyright infringement absent a license.

Last year's Supreme Court decision in Warhol Foundation v. Goldsmith blurred the relationship between commerciality and the first fair use factor, the "purpose and character" of the use. The litigation stemmed from events beginning in 1984, when famed artist Andy Warhol paid photographer Lynn Goldsmith a $400 licensing fee for the use of her black-and-white portrait of late musician Prince as an "artist's reference" in developing a series of 16 silkscreen prints for Vanity Fair magazine. The magazine published one of these prints in 1984, and published another commemorating Prince's death in 2016, paying only the Andy Warhol Foundation for the Visual Arts $10,250 for use of the image.

In a 7-2 majority opinion, the Court found that Warhol violated Goldsmith's copyright in the photograph, relying on the commercial use of Warhol's prints as a key factor in determining that Warhol had not engaged in fair use. Addressing only the "purpose and character" of the use, the Court overrode the contention that Warhol's alterations and additions to the original photograph were transformative, instead emphasizing that the use's purpose, commercial licensing to a magazine, was identical to Goldsmith's. In conflating commerciality with purpose and weighing the use's commercial purpose as the determining factor, the Court's opinion created uncertainty as to how "commercial" use and the "purpose" of a use will be defined and weighed against the other fair use elements by future courts.

Whether data scraping is fair use depends on commercial use. But Warhol made the line between what does or does not constitute commercial use unclear, as well as muddying the distinction between the first and fourth fair use factors (purpose/nature of the use and the market effect of the use, respectively) and how commerciality interacts with them. Thus, there is a concern that courts may use the Warhol decision to overwhelmingly latch onto commerciality as a deciding factor in upcoming AI fair use cases. Such an oversimplification in an increasingly complex arena may lead to even more confusion in the future.