The following are my remarks given on April 25, 2023 to the USPTO as part of their AI listening session:
by Dennis Crouch
Members of the USPTO, and fellow participants of this AI Listening Session, thank you for inviting me here today and for taking time to consider these important issues. I want to also thank the prior speakers who have done a great job laying out many of the issues. I am also happy to work with any of you to help figure this out and reach a workable system that truly encourages innovation.
My name is Dennis Crouch, and I am a law professor at Mizzou and author of Patently-O. It is my privilege to discuss the role of generative AI in the realm of intellectual property and the need for clear guidance from the USPTO.
As artificial intelligence progresses at an unprecedented pace, numerous cases have emerged where generative AI has played a crucial role in conceiving an invention. In certain instances, if the AI were human, it would be rightfully recognized as at least a joint inventor. This raises the question of whether it is appropriate to designate the human, who contributed to only a part of the invention and collaborated with the AI, as the sole inventor. This is particularly concerning in cases where the AI introduced concepts that the human had not conceived or even considered. Generative AI differs from traditional tools in that its responses are unpredictable and it produces results akin to those of a human inventor.
I would like to draw attention to a striking similarity between generative AI and biological models, particularly in the context of the pending Supreme Court case of Amgen v. Sanofi. In this case, researchers patented a genus of monoclonal antibodies, but the antibodies’ amino acid sequences were not designed by humans. Instead, a genetically modified humanized mouse generated the antibodies in response to a specific antigen. This scenario closely mirrors the role of generative AI in the invention process.
The parallel between the genetically modified humanized mouse and generative AI becomes apparent when we examine the prompt given to the mouse in the form of a PCSK9 injection and the subsequent response: antibodies collected from the mouse’s spleen. This analogy can be applied to generative AI, with a human providing the initial input or prompt, and the AI system generating an inventive output.
One key issue is the uncertainty patent attorneys face regarding the proper course of action. Innovators are seeking to protect their valuable inventions, but the lack of clear guidance creates potential ethical dilemmas for patent attorneys. I recently published an article on Patently-O titled “AI Inventor and the Ethics Trap for US Patent Attorneys,” highlighting this concern.
I believe the USPTO should promptly offer guidance, stating that patent applications may appropriately list the human contributor to the conception as the sole inventor, even in situations where an AI or other tool provided key elements of the discovery.
The US Copyright Office has taken steps to deny registration of AI-created works. It is essential for the USPTO to avoid the current pitfalls of the US Copyright Office in addressing AI-related issues. Ideally, US intellectual property policymakers would consider all aspects of IP—patent, trade secret, and copyright in our situation here—as a unified whole. This might present an opportunity to contemplate the establishment of a US Intellectual Property Office that merges the PTO and Copyright Office, while also providing some authority to regulate trade secrecy.
As several speakers have noted, generative AI is expected to reduce the cost of inventing, which is a tremendous benefit. As Profs Levine and Feldman explained, AI has different incentives than human inventors and lacks the fundamental humanity that our inventorship laws respect. Still, a valuable technological improvement by an AI (such as a new medical treatment) is something that we want to encourage.
For inventions without direct human contribution, it is timely to consider a special rights category for computer-generated inventions. This unique exclusivity could feature a reduced term and additional requirements to ensure clarity and patentability, such as pre-screening, limited claims, definitions, and the incorporation of born-digital aspects of the documentation.
In conclusion, it is imperative for the USPTO to provide guidance on how to handle generative AI’s role in the invention process – distinguishing between situations where a human inventor exists and those with in no or insufficient human originality.