Unpacking AI-Assisted Drug Discovery Patents

AI is vaulting drug discovery forward leaps and bounds—and now regulators are beginning to catch up, with the United States Patent and Trademark Office recently issuing new guidelines on the patentability of AI-assisted inventions.

For the latest webinar in a series of insights focusing on the topic, we analyzed what the new guidelines mean for the world of drug discovery.

Check out the webinar for our full insights. Here are some key takeaways:

Patents require a human inventor. But AI-assisted inventions are patentable if a human made a “significant” contribution. What constitutes a “significant” contribution isn’t explicitly defined—though the patent office has provided some guiding principles, and a clearer picture will develop through case law. Either way, regulators have made at least two things clear when it comes to AI-assisted inventions: There must be a human inventor on every claim in a patent, and the USPTO inventorship analysis will focus on human contributions, not whether AI was used in the conception.

Just following AI directions is (usually) not enough. In most cases, a human merely reducing an AI invention to practice—that is, taking the idea for an invention and manifesting it in the physical world—is not a significant enough contribution for the patent office. Iterating and improving on an AI output, however, can help clear their bar. There are some exceptions when it comes to drug discovery, though.

Doctrine of Simultaneous Conception and Reduction to Practice. When a molecule's synthesis is so unpredictable that it must be completed in a lab to establish its feasibility to be synthesized and isolated, the act of synthesizing an AI-generated molecule can be significant enough. For example, in the small molecule space, innovators probably still need a synthetic process to determine how to make the small molecule, and that can be deemed a significant contribution.

Specialized AI carries rewards—and potential risks. Building a specialized AI that’s designed to help find solutions for a specific problem can increase the patentability of the invention that was part of the AI’s output, because human ingenuity played a more significant role guiding AI toward that solution.

Ownership of AI Outputs. With increasing use of AI-enabled tools as SaaS or Platform as a Service-type products, customers should ensure ownership of AI outputs in agreements with providers where the provider personnel would likely be inventors. IP ownership terms should be carefully negotiated balancing the provider’s ability to provide services to third parties with the customer’s interest to protect pipeline IP.

Document human contributions. In this still-developing area of the law, it’s critical to thoroughly document the inventive process—both AI and human contributions—in case innovators encounter issues with regulators or a third party. Detail the specific acts humans undertook beyond mere verification of AI outputs. Document factual evidence establishing why the science is so unpredictable and highlight human contributions that lead to the successful invention.

Patent counsel is particularly adept at this kind of well-reasoned analysis designed to withstand enforcement and litigation years down the road. Engaging patent counsel with documentation helps preserve details that may otherwise be unintentionally destroyed through application of routine document-retention policies.

Conduct thorough patent due diligence. Don't merely rely on representations and warranties. It’s never been more important to conduct due diligence around AI inventorship. USPTO’s human inventorship thresholds are not clear and objective, and application to specific fact patterns will generate uncertainty. There is further uncertainty that courts may change USPTO’s guidance. However, given long litigation timelines, any adverse outcomes will in most cases fall out of the survival period of representation and warranties in transaction documents. Transactional partners are advised to conduct thorough due diligence to assess the risks around AI inventorship.