ARTIFICIAL INTELLIGENCE FAQs

A. Artificial intelligence (AI) is a “machine-based system with a component of machine learning or autonomous reasoning and/or decision making that can, for a given set of defined objectives, train on data and develop models to make predictions, recommendations, or decisions influencing real or virtual environments.” ABA Glossary. Machine Learning “leverages statistical algorithms to enable computers to perform tasks autonomously by learning from data. It is adept at identifying patterns and making informed predictions, with applications spanning spam detection, content recommendation, and advanced fields like voice and image recognition. Machine learning operates through distinct methodologies: supervised learning, where algorithms learn from labeled examples; unsupervised learning, which involves discovering patterns in data without explicit guidance; and reinforcement learning, where algorithms refine their actions based on feedback to achieve specific goals.” ABA Glossary, citing https://www.cloudflare.com/learning/ai/what-is-machine-learning/.

A. Generative AI refers to a type of AI “that can create new and original content, such as images, audio, text, or software code without being explicitly programmed to do so. This is achieved using deep learning algorithms that are trained on large datasets of existing content, allowing the AI to learn patterns and generate new content that is similar in style or structure.” ABA Glossary. Generative AI has given rise to some of the most complicated copyright questions (see below). Some common examples of Generative AI include text generation (i.e., ChatGPT, Gemini, Claude), image generation (i.e., DALL-E, Midjourney, Stable Diffusion), video generation and music and audio generation (i.e., ElevenLabs).

A. The Copyright Office has taken the position that “human authorship” is required for a work to be registered. Accordingly, a work which is entirely generated by AI and lacks human authorship is not eligible for copyright protection and is considered to be in the public domain. This position was affirmed by the U.S. Court of Appeals for the District of Columbia Circuit in March 2025 in Thaler v. Perlmutter, 130 F.4th 1039 (D.C. Cir. 2025). Where an applicant has made “de minimis” use of AI, that will not prevent registration. However, “appreciable” AI-generated content must be excluded from the application, leaving a gray area between “de minimis” and “appreciable.”

A. Yes, but the issue is complicated. (See our discussion of the partial cancellation by the Copyright Office of a registration granted for a graphic novel with AI-generated artwork.)

In its most recent Artificial Intelligence Report, published January 29, 2025, the Copyright Office stated that “the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements [emphasis added]. This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts. The Office confirms that the use of AI to assist in the process of creation or the inclusion of AI-generated material in a larger human-generated work does not bar copyrightability. It also finds that the case has not been made for changes to existing law to provide additional protection for AI-generated outputs.”

Similarly, Jason M. Allen was denied registration in September 2023 of a two-dimensional artwork which the Review Board of the U.S. Copyright Office found to contain “more than a de minimis amount of content generated by AI,” and Allen refused to disclaim the AI-generated content. In December, Ankit Sahni requested registration of a two-dimensional artwork created by an AI program, rendered in the style of Vincent van Gogh’s “The Starry Night.” The Board determined that “[w]hile Mr. Sahni selected the numerical variable for the ‘strength’ of the style, that choice alone is insufficient to warrant copyright protection . . .selecting a single number for a style filter is the kind of de minimis authorship not protected by copyright.”

In January 2025 the Copyright Office allowed the registration of an AI-generated artwork (“A Single Piece of American Cheese”) based on “selection, arrangement, and coordination of the AI-generated material,” as well as a visual collage (“A Collection of Objects Which Do Not Exist”) containing AI-generated elements.

The amount of AI involvement needed in order to render a work non-copyrightable remains fact-specific and often unclear. Advice of experienced copyright counsel in such cases is highly recommended.

A. For general background on patents, see our Patent FAQs.
The Patent Act requires all inventors to be an “individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.” 35 U.S.C. §100(f). The U.S. Patent and Trademark Office (“USPTO”) position that AI cannot serve as an inventor was unsuccessfully challenged by computer scientist Stephen Thaler. Thaler v. Vidal, 43 F.4th 1207, 1212 (Fed. Cir. 2022) (“[T]he Patent Act, when considered in its entirety, confirms that ‘inventors’ must be human beings.”) That being said, AI-assisted inventions are patentable in cases where a natural person provided a “significant contribution.” It should also be noted that Thaler’s AI system was granted patents in South Africa, one indication that patentability of AI-created inventions varies by country. Thaler, 43 F.4th at 1212.

A. Trade secrets may be a more promising way to protect AI creations than patents since trade secret law does not currently require human involvement in the creation of secret information. Additionally, a trade secret does not need to meet the utility requirement. While trade secret protection could in theory extend to every aspect of generative AI, there will be challenges both in identifying with specificity the algorithms and data sets that constitute the trade secrets and in protecting them from unauthorized disclosure.

A. Owners of content in many media are alleging that use of their works to train AI constitutes infringement of their IP rights. The lawsuits to date have been brought by the creators of books, newspapers, online news media, musical works and sound recordings, visual works, photographs, YouTube videos, voice actor recordings, and even Westlaw case reports, and generally involve allegations of copyright infringement.

While each lawsuit alleging copyright infringement is fact-specific, there are common themes. For example, AI developers and researchers train AI systems with datasets containing content from websites across the internet, which may contain copyright-protected material. AI-generated content may also infringe on copyright-protected material if it produces a new work including elements of a work owned by someone else (i.e., prompting AI to create a work that draws on a character protected by copyright).

Some authors of copyrighted material have also alleged that use of their content results in the removal of copyright management information (“CMI”) in violation of the Digital Millennium Copyright Act. Social media users have also brought suits alleging misuse of their personal and copyrighted data for purposes of training AI systems.

One of the most notable recent cases alleging the use of copyrighted information for purposes of training AI models is New York Times Company v. Microsoft Corporation, in which the newspaper alleges that Open AI and Microsoft have used millions of its copyrighted articles without permission to train the large language model behind AI tools like ChatGPT. In April 2025 the U.S. District Court for the Southern District of New York issued a preliminary ruling allowing most of the plaintiff’s claims to continue. NY Times Co. v. Microsoft Corp., 2025 U.S. Dist. LEXIS 64462.

Professor Edward Lee of Santa Clara University keeps a comprehensive running list of the copyright lawsuits brought against AI companies.

A. In addition to the usual provisions in a software licensing agreement, licensees will want to pay particular attention to the intellectual property and data privacy representations and warranties as to non-infringement. Ideally any limitations on the licensor’s liability should exempt breach of these key provisions.

The concept of collective licensing, which exists for use of copyrighted music, offers a potentially simple and effective way to address the use of copyrighted material in the creation of AI systems.