September, 17th, 2019

By Mackenzie L. Long

Earlier this year the United States Patent and Trademark Office (USPTO) released eligibility guidelines for AI-related inventions. However, there still remains multiple questions regarding the ability to patent AI inventions. Late last month the USPTO published a request for public comments to determine whether further examination guidance is needed to “promote the reliability and predictability of patenting artificial intelligence inventions.” Request for Comments on Patenting Artificial Intelligence Inventions, 84 Fed. Reg. 44889, 44889 (Aug. 27, 2019).

The Request for Comments includes twelve questions to help “aid the USPTO in collecting relevant information to evaluate whether further guidance is needed and assist in the development of any such guidance with respect to patenting artificial intelligence inventions.” These questions range from inventorship and ownership to written description, enablement, and subject matter eligibility.

Arguably the most important question is that of ownership/inventorship. Currently the USPTO only recognizes humans as inventors. Now the USPTO is asking how inventorship for AI-based inventions and inventorship for inventions discovered by AI should be defined.

In the US an inventor is any person who contributed to the conception or reduction to practice of an invention claimed in a patent application. However, when a new AI-based invention is created, the inventorship could be given to not only the computer scientist who conceived the algorithm used by the AI, but also the scientist who trains the AI, and the engineer who identifies a real-world application for the AI.

Even more complicated are inventions discovered wholly or partially by AI. This can be illustrated by a new computer program that can autonomously design synthetic drug routes to navigate around patent-protected syntheses of blockbuster drugs.

Researchers Karol Molga, Piotr Dittwald, and Bartosz Grzybowski, in Poland and South Korea, have unveiled computerized methods to suggest synthetic strategies that bypass patent-protected aspects of essential drugs, called Chematica. Chem, Molga et al. “Navigating around patented routes by preserving specific motifs along computer-planned retrosynthetic pathways.” To get unpatented options, Chematica keeps track of lists of specific bonds a scientist wishes to preserve, and identifies the key disconnections used in the patented process and designs synthetic routes that circumvent the patented approaches.

“By algorithmically locating the key bonds on which patents hinge and propagating them down Chematica’s retrosynthetic trees, we can generate synthetic solutions from alternative yet economical starting materials, achieving a real practical impact.”

Chematica will help generic drug companies create more efficient and economical syntheses of drugs, driving down once expensive and exclusive medication. However, for companies who use Chematica to develop new drugs and try and patent large numbers of synthetic routes, there is the question of who is the inventor and owner of the new drugs and syntheses. Is it the drug company and their scientists? Or is it Molga, Dittwald, and Grzybowski, who own Chematica?

This is an important opportunity for individuals and companies active in AI technologies, including those in the pharmaceutical industry, to provide their perspective on how U.S. intellectual property law and patent office procedures should address these issues for AI inventions. Written comment may be sent by email to, and must be received on or before October 11, 2019.


WHIPgroup is leading counsel for U.S. and international technology companies. We specialize in patent and trademark law.

© 2019 Whitmyer IP Group, Stamford, Connecticut.

© Copyright 2024 Whitmyer IP Group