March 12, 2024

USPTO’s New Guidance on AI-Assisted Inventions: The Impact on the Use of AI in the Life Sciences


On February 12, 2024, the US Patent Office and Trademark Office (USPTO) released the Inventorship Guidance for AI-assisted Inventions (the Guidance). We previously discussed the Guidance here.

Following up on the Guidance, the USPTO released two examples illustrating what the USPTO considers proper inventorship analyses for AI-assisted inventions. Each example sets forth different fact patterns and walks through an analysis of whether one or more human individuals qualify as inventors. Acknowledging that life sciences companies are increasingly employing AI systems to help identify molecular targets and/or design therapeutic molecules, one of the two examples focuses on the use of AI to develop therapeutic molecules: Developing a Therapeutic Compound for Treating Cancer (Example 2).

Life sciences companies using AI-assisted systems should carefully consider whether their current R&D efforts allow for natural persons to provide a significant contribution such that the resulting efforts may properly identify a human inventor.

USPTO Guidance

AI-assisted inventions are not categorically unpatentable
The Guidance starts by establishing that AI-assisted inventions are not categorically unpatentable. It then notes that the patent system is set up to “incentivize and reward human ingenuity,” and as such, patent protection is limited to “inventions for which a natural person provided a significant contribution to the invention.”

Significant contribution is determined using the Pannu factors
In establishing whether an individual is an inventor or joint inventor, the Guidance focuses on three determining factors (the Pannu factors). Notably, each of the three factors should be satisfied to establish an individual as an inventor or joint inventor.

Per the Guidance, an individual is an inventor if the individual:

  1. Contributes in some significant manner to the conception or reduction to practice of the invention
  2. Makes a contribution to the claimed invention that is not insignificant in quality, when that contribution is measured against the dimension of the full invention
  3. Does more than merely explain to the real inventors well-known concepts and/or the current state of the art

Guiding principles for determining inventorship
The Guidance further sets forth a nonexhaustive list of guiding principles. Portions of the guiding principles appear below:

  1. A natural person’s use of an AI system in creating an AI-assisted invention does not preclude a person from being an inventor if the natural person significantly contributed to the invention.
  2. Merely recognizing a problem or having a general goal or research plan to pursue does not rise to the level of conception. A natural person who only presents a problem to an AI system might not be a proper inventor of an invention identified from the output of the AI system.
  3. Reducing an invention to practice alone is not a significant contribution that rises to the level of inventorship. Therefore, a natural person who merely recognizes and appreciates the output of an AI system as an invention, particularly when the properties and utility of the output are apparent to those of ordinary skill, might not necessarily be an inventor.
  4. A natural person who develops an essential building block from which the claimed invention is derived may be considered to have provided a significant contribution to the conception of the claimed invention even though the person was not present for or a participant in each activity that led to the conception of the claimed invention. In some situations, the natural person(s) who designs, builds, or trains an AI system in view of a specific problem to elicit a particular solution could be an inventor, where the designing, building, or training of the AI system is a significant contribution to the invention created with the AI system.
  5. Maintaining “intellectual domination” over an AI system does not, on its own, make a person an inventor of any inventions created through the use of the AI system.

Developing a Therapeutic Compound for Treating Cancer (Example 2)
The implementation of AI and machine learning (ML) is a valuable strategy for many innovators in the life sciences. For example, ML models can be trained and deployed to more effectively identify binders to target proteins and/or design new drugs or therapeutics. Example 2 describes two life sciences R&D scenarios using AI.

The tables shown below break down the two scenarios presented in Example 2 and the corresponding takeaways.

Example 2, Scenario 1

A “ready-to-use” deep neural network (DNN) model is used to predict binding affinity between drug-target pairs. The DNN predicts the binding affinity of input compounds for a particular receptor (mutated androgen receptor protein). The compound with the highest affinity (CID_1) is synthesized and modified by a human individual to improve selectivity (CID_1-mod). The applicant files claims to (1) a method of identifying, synthesizing, and modifying compounds using the DNN; and (2) DID_1-mod.

This scenario implicates guiding principles 2, 3 and 5 of the Guidance.

USPTO Examples Inventorship Conclusion Takeaway
Lauren trains a deep neural network (DNN) model on diverse sets of compounds and targets from previous drug-target experiments. Lauren did not design the model with the specific problem in mind. Lauren maintains the DNN model.
Lauren is not an inventor to a claimed method of identifying and synthesizing a lead drug compound, nor to a claimed composition of the CID_1-mod.
Building a general AI/ML model and generally maintaining the model is not a significant contribution to an invention that solves a specific problem (e.g., binders to a mutated androgen receptor protein).  
Raghu uses the “ready-to-use” DNN model to predict drug compounds with high binding affinity to a mutated receptor. To use the DNN model, Raghu inputs a target receptor and several candidate drugs. The DNN outputs a numerical value representative of the binding affinity for each drug to the receptor. Raghu sorts the DNN model output in descending order.
Raghu is not an inventor to a claimed method of identifying and synthesizing a lead drug compound, nor to a claimed composition of the CID_1-mod.
Taking an off-the-shelf AI/ML model and applying it to a problem may not be a significant enough contribution to be an inventor.
Marisa and Naz synthesize compounds predicted by the DNN to have the highest affinity. They further perform structural modifications to increase binding selectivity, creating a novel therapeutic drug compound, CID_1-mod.

Marisa and Naz are joint inventors of a claimed method of identifying, synthesizing, and modifying a lead drug compound.

Marisa and Naz are joint inventors of the claimed CID_1-mod.

Modifying the output of an off-the-shelf AI/ML model may rise to a significant contribution to establish inventorship.

Example 2, Scenario 2

In scenario 2, two human individuals contribute to developing and building a new generative neural network-based AI system referred to as molecule optimizer (MO), which creates new molecules taking into account specific desirable properties. The human individuals identified problems with drugs predicted using a previous AI model. They set out to find a particular solution to these problems. The MO was designed by the human individuals to incorporate five specific desirable absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties for potential therapeutic compounds. The human individuals trained and fine-tuned the MO on specific datasets relevant to the desired analysis. The human individuals employ a “ready-to-use” DNN model to identify initial drug candidates. The human individuals then feed the DNN-identified candidates through the MO to create a structurally modified compound (MID_1), which meets the desired ADMET properties. The applicant files a claim to the structurally modified compound (MID_1).

This scenario implicates guiding principles 1 and 4 of the Guidance.

USPTO Examples Inventorship Conclusion Takeaway
Raghu develops the molecule optimizer (MO) and trains it on specific datasets. Marisa identifies desirable properties of drug compounds for use in training the MO. The contributions of Raghu and Marisa result in the creation of the MO, which overcomes specific problems in a prior program. The MO in turn optimizes drug compound structures and predicts a novel compound MID_1.
Raghu and Marisa are both inventors of a claimed composition of the novel compound MID_1.

A significant contribution necessary for inventorship may arise when an individual builds and trains a new AI/ML model to solve a specific problem.

Contemporaneously recognizing and appreciating an invention of an AI may complete conception.


This informational piece, which may be considered advertising under the ethical rules of certain jurisdictions, is provided on the understanding that it does not constitute the rendering of legal advice or other professional advice by Goodwin or its lawyers. Prior results do not guarantee a similar outcome.