AI Drug Discovery Credits Spark Patent Dispute

Insilico Medicine's press release last year had no trouble crediting its generative AI platform for the design of a new drug candidate for pulmonary fibrosis, calling it "discovered by" the software. The patent filing told a different story. Five humans, including CEO Alex Zhavoronkov, are listed as the inventors. No algorithm appears anywhere on the document.

That gap between marketing and paperwork isn't an oversight. It reflects a hard line in patent law: only people can be named as inventors, no matter how much of the actual chemistry an AI system worked out on its own. Insilico is far from alone in leaning on machine learning to generate drug candidates that no chemist might have proposed unprompted, and the approach is spreading fast across the industry as models get better at spitting out atomic structures on demand.

The rule barring AI from inventorship credit didn't come from nowhere. Ryan Abbott, a partner at the Los Angeles firm Brown, Neri, Smith & Khan, tested it directly in US courts on behalf of an AI system called DABUS, arguing that the software deserved sole credit for a cleverly designed food container. That case forced judges to answer a question patent law had never had to confront before.