Last week I had the honor of presenting at the International Trademark Association (“INTA”) conference on The Business of AI with my distinguished co-panelists. Professionals from various sectors including academics, business, and law attended. In addition to attorneys, representatives from AI firms, consultancies, and a creative agency presented their unique perspectives on AI’s application to the intellectual property (“IP”) field.

My session focused on the implications of current and pending U.S. regulations regarding AI, and my co-panelists focused on international (EU and Asia) laws. We all discussed emerging efforts by the federal government and domestic agencies such as the Copyright Office and United States Patent and Trademark Office (“USPTO”) to regulate AI. I wanted to share some key takeaways about where the U.S. stands on regulating AI in the IP context:

President Biden’s Executive Order of October 30, 2023

The Executive Order, which I noted in a previous post, offers a good framework for characterizing the risks associated with AI and delegates the mitigation of these risks to various governing bodies. Timing is key in the Order for forthcoming regulation by the USPTO and Copyright Office. Such regulations will be triggered by the issuance of a report by the Copyright Office to be delivered in segments during 2024. As the report comes out, the USPTO and Copyright Office will consult on forthcoming regulations governing use of AI in the context of patent, trademark, and copyright law.

Copyright Office

The Copyright Office has found that works generated solely by machine (that is, AI) are not registrable for copyright, although it has recognized that human authorship may exist in contributions to the text or compilation of a work.

Court Cases

Whether the training and output of AI platforms is a copyright infringement will likely depend on the upcoming decisions of cases like New York Times Co. v. Microsoft Corp. (which I’ve reported on here and here) and Thomson Reuters v. Ross Intelligence. In both cases, news publications are claiming that AI training and generative output infringe copyright in their journalism. In the meantime, social media sites are likely to follow Reddit in reaching settlements with AI platforms (in Google's case, Gemini) to train on the social media sites’ content, in exchange for hefty compensation ($60 million in the case of Reddit).

Challenges in Proving Copyright Infringement by AI

Regulation of AI for instances of copyright infringement may become more complex than expected. This is because the various forms of AI, such as transformer versus diffusion-based models, could call for a closer look at how exactly source material is being used and if the content is actually infringing. As of now, researchers are not able to discern exactly which source material is used for generating output that could cause difficulties for parties alleging their works were used and infringed upon.

State Regulation of Deepfakes

There has been a strong state-led response to impose regulations on deepfakes. The most notable is the ELVIS Act in Tennessee, but many other states are working on legislation to protect artists’ publicity rights while still allowing for First Amendment exceptions. Many existing state laws—at least ten states to date— regulate political campaign and nonconsensual, pornographic deepfakes.

AI Tools in the Private Sector

AI will likely become a productive tool for the private sector. While companies are limiting their use of general AI devices like ChatGPT, devices that are specifically tailored to help professionals are emerging rapidly in the market and will likely be encouraged in the interest of efficiency.