South SFV Today

In a recent study published in Nature Biotechnology, researchers from Stanford University, including Professor Lisa Larrimore Ouellette from Stanford Law School, examined the impact of artificial intelligence on patent disclosures. The paper addresses how AI's role in drafting patents may affect the clarity and usefulness of technical literature for future researchers and innovators.

The authors highlight that patent law requires inventors to describe their inventions clearly enough for others skilled in the field to replicate them. This disclosure is crucial for disseminating knowledge and ensuring that patents cover only what inventors have genuinely contributed. However, they note that enforcing these standards has been inconsistent due to time constraints and limited training among patent examiners.

"AI will amplify errors in both improperly granting poorly disclosed patents and improperly denying patents on the basis of earlier AI-generated references that do not actually disclose those inventions," write co-authors Lisa Ouellette, Victoria Fang, and Nicholas T. Ouellette.

Despite potential issues, the authors see AI as an opportunity to reform the patent examination process. "The disruption to the patent drafting and review process created by AI gives us an opportunity to improve disclosure across the patent system," said Lisa Ouellette.

The United States Patent and Trademark Office (USPTO) has issued guidance requiring human review of AI-drafted applications. However, the authors argue this is insufficient and suggest raising disclosure standards. They recommend requiring inventors to implement some version of their invention before filing or meet stricter requirements for unproven technologies. Improved training for patent examiners is also advised.

To assess AI's capabilities in drafting disclosures, the researchers evaluated two patent-specific AI systems (Edge and Vaero) and ChatGPT-4o with technologies related to fluid dynamics. For existing technologies, AIs produced detailed specifications but included inaccuracies alongside correct information. For hypothetical technologies, AIs failed to generate workable solutions yet produced seemingly credible but incorrect specifications.

"Interestingly, the AI-generated specifications for both existing and hypothetical technologies were often sufficient to satisfy the current level of scrutiny provided by patent examiners even if they did not meet scientific standards," said co-author Fang.

"The outputs of AI tools can look convincing at first glance but can also have serious flaws," noted Nicholas Ouellette. "Without careful review, they risk diluting the value of patent disclosures as a source of technical knowledge."

Lisa Ouellette emphasized that while AI tools show promise in reducing costs and inequalities in accessing the patent system, they reveal deeper issues with disclosure standards enforcement.

"By refining disclosure standards, reconsidering examination practice, and using technology wisely, we can ensure that patents are granted only for what the inventor actually invented and disclosed," she stated.

This research was originally published by Stanford Law School.