The recent surge of accessible generative AI (“GenAI“) tools has kept attorneys, particularly those in the intellectual property, technology, data privacy, and cybersecurity spaces, on their toes. Within the intellectual property community, there have been ongoing discussions, incremental resolutions, and evolving opinions. In this article, we will outline and summarize the current intellectual property legal landscape in the United States related to GenAI.

I. OVERVIEW

In broad terms, GenAI tools work by digesting large amounts of data (referred to as “inputs”), learning patterns, structures, and relationships within the inputs along the way. GenAI tools use the patterns, structures, and relationships learned to generate “new” content (referred to as “outputs”). Outputs may include images, stories, sounds, software code, and more.

Courts, as well as intellectual property professionals and organizations, have been grappling with, among other things: (1) whether GenAI outputs are patentable or copyrightable and, if so, who owns the rights; and (2) whether using protected works to train GenAI tools infringe upon the original owners’ intellectual property rights.

II. DECISIONS AND GUIDANCE

A. Protection & Ownership

In August of 2023, the U.S. District Court for the District of Columbia reaffirmed an earlier decision finding a GenAI output, without sufficient human contribution, is not copyrightable. This finding is in line with guidance issued by the U.S. Copyright Office (“Copyright Office“). The extent of human contribution required to merit protection, however, is still up for debate. Regardless, the Copyright Office has issued further guidance stating the AI-generated portions of a copyrightable work will remain unprotected. Thus, from a copyright perspective, GenAI outputs are not protectable and not owned by anyone.

The U.S. Patent and Trademark Office (“USPTO“) allows for AI-assisted inventions to obtain patent protection, but a human must be significantly involved in developing the invention. The USPTO’s stance was reinforced in August of 2022 by the U.S. Appellate Court for the Eastern District of Virginia, which found an AI system cannot be an inventor of a patent. This result is different from copyrights because the ultimate invention is protectable, including AI-assisted portions, whereas the portions of a work generated by AI are not copyrightable.

B. Infringement

Infringement lawsuits brought against companies operating GenAI tools primarily involve copyright infringement claims. Such copyright infringement claims include direct infringement (using copyrighted materials as inputs without consent and creating unauthorized derivative works) and vicarious infringement (via the activities of the users of the GenAI tools). There are several notable, pending copyright infringement cases which are anticipated to provide much needed insight on how copyright infringement claims involving GenAI will be handled.

However, some initial rulings within these pending cases are already worth noting. For example, the U.S. District Court for the Northern District of California (the “Court“) recently ruled a plaintiff’s GenAI prompts, outputs, and associated settings were not protected by the work product doctrine when tested in preparation for a copyright infringement lawsuit and included in a complaint. This ruling provides some insight on the discoverability of GenAI prompts, outputs, and settings in copyright infringement lawsuits involving GenAI.

In Kadrey v. Meta Platforms, Inc., the Court rejected plaintiff’s assertions that all outputs generated through an AI system infringe the copyrighted works used to train the system, indicating something more than mere use of copyrighted works to train AI systems is required to find copyright infringement based on outputs. Additionally, in Andersen et al. v. Stability AI Ltd., the Court dismissed a vicarious infringement claim because there was no claim of direct infringement.

While initial rulings have provided some indication of what is to come, the handling of copyright infringement claims stemming from GenAI is still largely unknown.

C. Other Government Guidance

On October 30, 2023, President Biden signed an Executive Order regarding AI. The Executive Order represents a sweeping mandate to guide responsible AI development and use. Notably, the Executive Order called for: (1) the Copyright Office to publish an AI study to address copyright issues raised by AI; and (2) the USPTO to publish guidance to patent examiners and applicants addressing inventorship and use of AI in inventive processes. For more information about the Executive Order, read our prior article: Decoding the White House’s Executive Order on AI.

i. Copyright Office

The Copyright Office published Part One of the highly-anticipated Report on Copyright and Artificial Intelligence on July 31, 2024 (“Part One“) pursuant to the Executive Order. Part One focuses on the legal and policy issues related to digital replicas—a video, image, or audio recording which has been digitally created or manipulated to realistically, but falsely, depict an individual (also known as a “deepfake”). Part One details the harms of digital replicas, as well as the existing legal frameworks to protect against the unauthorized digital replicas and imitations of human artistic style. Importantly, Part One includes a call upon Congress to enact a federal law protecting individuals from unauthorized deepfakes.

After the Copyright Office published Part One, California governor, Gavin Newsom, signed AB. 2839, AB 2655, and AB 2355 into law in order to address the use of political deepfakes. AB 2839 is of particular importance because it went into effect immediately and prohibits certain uses of materially deceptive audio or visual media of political candidates without appropriate disclosures. AB 2655 will go into effect on January 1, 2025 and requires social media platforms to block or label materially deceptive content related to elections. AB 2355 will also go into effect on January 1, 2025 and requires electoral advertisements using GenAI or substantially altered content to contain certain disclosures.

An individual filed a lawsuit against California on the same day AB 2839 went into effect, alleging AB 2839 and AB 2655 violate the First and Fourteenth Amendments and seeking a preliminary injunction to block enforcement of AB 2839. The U.S. District Court for the Eastern District of California found AB 2839 likely violates the First Amendment and granted the preliminary injunction. This decision and the ultimate outcome of the case could significantly hinder regulation of deepfakes.

ii. USPTO

The USPTO issued its Guidance Update on Patent Subject Matter Eligibility, Including Artificial Intelligence on July 17, 2024 (“Patent Guidance“) pursuant to the Executive Order. As indicated by the title, the Patent Guidance provides additional guidance on determining subject matter eligibility of AI inventions under patent law. The Patent Guidance notably confirms patent protection may be sought for AI-assisted inventions where one or more persons made a significant contribution to the claimed invention. The Patent Guidance also contains examples intended to guide and clarify application of the Alice/Mayo test for analyzing subject matter eligibility as related to AI inventions.

III. TAKEAWAYS

As intellectual property law is applied and interpreted by courts, the Copyright Office, USPTO, and others in the AI context, it is important for businesses developing and using GenAI to stay abreast of the latest rules and guidance. We plan to release new articles and blog posts as the courts, government entities, and others continue to provide clarity.