As AI tools become more prevalent in the life sciences, biotechnology companies need to evaluate AI’s impact on their ability to protect the results of their research. Recent U.S. Patent and Trademark Office (PTO) guidance can help inform your scientific development process, from experimental design to patent prosecution to legal disputes that may arise.
The big takeaways from the guidance are:
- Claimed inventions must have a human inventor. AI cannot be an inventor.
- Using AI does not preclude granting a claim, so long as a human made a contribution sufficient to be named as an inventor.
- In determining whether a human made an inventive contribution, the standard legal factors used for determining inventorship apply.
You can treat AI in an inventorship analysis as if it were any other member of the inventor group, but you cannot get a patent if the AI is the only inventor.
Based on PTO guidance from February and July, biotech companies should consider the following strategies to help ensure patentable protection for AI-supported innovations:
- Ensure humans are involved to identify “a specific problem to elicit a particular solution.”
- Have human decision-makers make a “substantial contribution” to identifiable claim elements and get your patent lawyer involved early.
- Consider the inventive contributions of those creating or fine-tuning AI tools.
- In publications or press releases, emphasize the human involvement in conception of the claimed invention and in directing the AI tool’s process to achieve the desired outcome.
- Prepare for patent eligibility challenges.
In More Detail: Ensuring Patentable Protection for AI-Supported Inventions
1. Ensure humans are involved to identify “a specific problem to elicit a particular solution.”
For any part of an invention to be entitled to patent protection, it must have been “conceived” by a human inventor. The human must do more than merely recognize a general problem to be solved.
A solution that emerges by using a broad AI tool with generalized instructions may not be entitled to patent protection. As a result, it is important to show that humans had a specific problem and a particular solution in mind when designing or prompting the AI tool. A human building a specialized AI tool to solve a particular problem can form the basis of human inventorship of the derived solutions.
In drug discovery, for example, aim to have humans make framing decisions in the beginning of the process, such as identifying the class of compounds to research. This is one example of humans having a “particular solution” in mind so that the resulting invention could be considered “conceived” by a human.
To cite another example, if you use AI tools to count or classify cells, make sure a human decision-maker identifies the type of cells or cell markers.
2. Have human decision-makers make a “substantial contribution” to identifiable claim elements and get your patent lawyer involved early.
Every claim must have some inventive human contribution. Because inventions often involve a combination of multiple steps or components, track where humans contribute to the conception of specific claim elements. This may require planning your patent-claim drafting strategy during AI development. You may want to engage your patent lawyer earlier in the R&D process than you have in the past.
3. Consider the inventive contributions of those creating or fine-tuning AI tools.
Any person “who designs, builds, or trains an AI system” that aims to solve a specific problem in a particular way might be considered an inventor of the resulting invention if their contribution is significant enough. These individuals will need to be listed as inventors if you pursue a patent application. Identifying the wrong inventors in an application makes a patent vulnerable to challenges.
Given the potential importance of tool designers to inventorship in AI-based inventions, take care with third-party vendors that design AI tools for you. Ensure contracts with vendors include a present assignment of all relevant IP rights. The contracts also should include the ability to have the vendor provide additional information and documentation that may be required to prosecute and/or defend those IP rights.
4. In publications or press releases, emphasize the human involvement in conception of the claimed invention and in directing the AI tool’s process to achieve the desired outcome.
It is important to discuss your AI tools consistently in communications with the PTO and the public. Make sure external communications also acknowledge the influence and guidance of the people who built these tools. These publications and discussions could influence the way that the PTO interprets the inventorship of your patent.
The PTO has not articulated any substantive change to disclosure requirements related to AI-based inventions, but you should prepare for that eventuality. Maintain records that identify which portions of the claims were conceived by humans versus those that came from the AI.
5. Prepare for patent eligibility challenges.
AI-derived inventions have been a focus of potential subject-matter eligibility challenges.
The PTO awards patents for “process[es], machine[s], manufacture[s], and composition[s] of matter.” The PTO will not award a patent for an application that describes “abstract idea[s], law[s] of nature, or natural phenomen[a]” without providing additional elements that add a technological improvement to the existing standard. The additional claim elements can’t be well-understood, routine, or conventional.
The PTO’s July guidance provided examples of AI-based patent claims that may describe patentable improvements. The PTO’s analysis of these examples emphasizes the importance of including claim limitations to specific tangible details to help avoid challenges.
For example, the PTO analyzed a hypothetical AI-based invention that used an AI model to analyze glaucoma surgery patients to identify risk of post-implantation inflammation. The PTO’s analysis found a claim ineligible that focused on merely collecting a data set, identifying patients at risk using a model trained on the data set and prescribing a treatment. However, performing treatment with a specific “Compound X” was sufficient to render the claim eligible because it combined a treatment specific to a patient population that AI modeling determined to be high risk.
This example also shows a way to recognize human contributions in AI-related inventions. While the particular selection of in-need patients may not have been a human conception, especially if the training on the data set was highly automated, using that information in a complete patent eligible invention that formed a specific treatment is something that likely would require a human.
