Where do US Copyright laws currently stand in regards to AI?

The current status is quite literally defined as “to be determined”, even by the United States Copyright Office. 

After notable findings this summer, on August 30th, the Copyright Office has issued a notice stating that it is  seeking insight on “the use of copyrighted works to train AI models, the appropriate levels of transparency and disclosure with respect to the use of copyrighted works, and the legal status of AI-generated outputs”. This is both a fact-finding endeavor to inform its own study, but also a vehicle to assess  whether legislative or regulatory steps are warranted.  
For creators of any type of copyrighted work (such as authors of books, online courses, photographers, etc), how much AI is a permissible contribution in order to copyright the work? Right now, the legalities around AI and in particular copyrights, are so complicated, there is no straightforward answer. As the US Copyright Office (USCO) receives an increasing number of applications containing AI-generated material, the USCO is as motivated to find an answer to this question as we business owners are.

Copyrights are a type of intellectual property that protects original works of authorship as soon as an author fixes the work in a tangible form of expression. Therefore, it’s easy to see why this legal quagmire has become so complex- when AI is used to create a work, who is the actual “author”, or “creator” of the work? How much should courts factor in “outputs” v. “inputs” when determining originality?

In its notice of inquiry, the USCO stated that “over the last year, AI systems and the rapid growth of their capabilities have attracted significant media and public attention,” with the adoption and use of generative AI systems by millions of Americans – and the resulting volume of AI-generated material – sparking “widespread public debate about what these systems may mean for the future of creative industries and raise significant questions for the copyright system.” 

While the USCO is attempting to answer approximately 30 questions, this notice primarily attempts to clarify:

• How copyright issue relate to training and the results produced by AI (ie, “output v. input”)
• When does copyrighted work, used without permission to train AI models, fall within the purview of fair use laws?
• How can AI ethically collect information to answer the above?
• When can someone use copyrighted works without permission to train AI models, and when is that ok with fair use laws?
• Are changes necessary to Copyright Act itself to further clarify when human involvement is necessary for copyright protection, especially in cases involving AI-generated content?

(Written comments are due by October 18, 2023, according to the notice, and written reply comments are due on November 15, 2023). We will provide an updated addendum to this article as soon as more information is available. 

The USCO is attempting to specifically define: 

1. The use of copyrighted works to train AI models. The USCO states that it is “aware that there is disagreement about whether or when the use of copyrighted works to develop datasets for training AI models (in both generative and non-generative systems) is infringing.
2. The proper scope of copyright protection for material created using generative AI. The Office states that “although we believe the law is clear that copyright protection in the United States is limited to works of human authorship, questions remain about where and how to draw the line between human creation and AI generated content. For example, are there circumstances where a human’s use of a generative AI system could involve sufficient control over the technology, such as through the selection of training materials and multiple iterations of instructions (“prompts”), to result in output that is human-authored? Resolution of this question will affect future registration decisions.
3. How copyright liability principles could apply to material created by generative AI systems. For example, if an output is found to be substantially similar to a copyrighted work that was part of the training dataset, and the use does not qualify as fair, how should liability be apportioned between the user whose instructions prompted the output and developers of the system and dataset?
4. The treatment of generative AI systems’ ability to mimic human voices, likenesses, or styles. Although these personal attributes are not generally protected by copyright law, their copying may implicate varying state rights of publicity and unfair competition law, as well as have relevance to various international treaty obligations. This is a unique question that would not typically be considered by the USCO, but has been raised based upon the overlap of questions inherently raised within copyright issues. 

Additionally, the USCO referenced a few other notable questions, including: 

•  Generative AI Labeling: Should AI-generated content be disclosed, or labeled as such? The USCO touches on this very relevant question, stating “Should the law require AI-generated material to be labeled or otherwise publicly identified as being generated by AI? If so, in what context should the requirement apply and how should it work?” (An appropriate addition given Google’s recent announcement that it has created a process by which it can permanently and invisibly mark images that will identify them as computer-generated in order to help prevent the spread of misinformation.)
• Right of Publicity: A number of individuals have testified before the Senate Intellectual Property Subcommittee, in an attempt to answer the question of the federal right of publicity causes of action, particularly as they’ve come to light in association with AI-generative models. To this point, the USCO asks “What legal rights, if any, currently apply to AI-generated material that features the name or likeness, including vocal likeness, of a particular person?” 
• Finally, the USCO references two key Supreme Court decisions that relate to the issue at hand (Google v. Oracle America and Andy Warhol Foundation v. Goldsmith). To this point, the USCO asks how should the ‘purpose and character’ of the use of copyrighted works to train an AI model be evaluated? What is the relevant use to be analyzed? Do different stages of training, such as pre-training and fine-tuning, raise different considerations under the first fair use factor?” 

As business owners, we are experiencing a rare, true, “wild west” moment in intellectual property law. We have the ability to work more efficiently and effectively with AI, but at the same time, are completely hamstrung by the fact that we do not know what we will be able to own, that we think we create. It’s tricky, but at the same time, legislation is indicating that an answer will eventually be clarified, leaving us with black and whites, instead of this gray zone. For the time being, the answer we are implementing within the firm, and what we are telling our clients: if you are making the effort to adopt and incorporate AI your business, for the time being, the best practices are to take it one step further, and ensure that you are truly creating; not just implementing a plug and play into your content or creations.