AI is poised to revolutionize drug discovery, but uncertainty in developing protectable IP in this emerging field creates a host of potential risks for companies innovating in this space. Further complicating things: not all AI-assisted compounds are created equal—and that extends to their patentability.
IP risk varies depending on the class of drug involved—the easier and more predictable a compound is to synthesize, the greater the potential risk.
When it comes to patenting AI-assisted inventions, the U.S. Patent and Trademark Office (USPTO) has said that a human must make a “significant” contribution to the “conception” of a new drug—but instantiating that concept into a real-world compound (what the law calls “reduction to practice”) is not enough to confer inventorship.
For some compounds, where synthesis methods are well understood and predictable, “conception” for patent purposes might occur as soon as AI generates the compound’s digital representation. For more complex and unpredictable compounds, a legal doctrine called “simultaneous conception and reduction of practice” applies.
Under the doctrine, conception may not be completed until researchers can demonstrate that they have successfully isolated the compound in the wet lab. So, the stage at which conception is deemed complete depends on how difficult it is to convert an AI-generated digital representation of a compound into a physical sample.
Antibodies and polypeptides are examples of drug classes that may carry higher IP risks if an unaltered AI-generated sequence turns out to be the final drug candidate. The synthesis methods of some of those amino acid sequences, particularly simpler ones, are quite well-known and standard that a researcher would send digital sequences to third-party vendors to synthesize. Since such sequences are easily synthesized once generated, in the future, courts may hold that AI alone completed conception if the AI-generated digital sequence corresponds to the subject matter of a patent.
This presents pharmaceutical companies with a dilemma when deciding whether to adopt an AI tool for discovering antibody/polypeptide drugs. A polypeptide’s potential variations are virtually infinite, so AI is well-suited to screen and simulate a large number of polypeptide variations. However, the moment AI outputs a viable polypeptide sequence, conception may have been immediately completed.
A pharmaceutical company could face a tough decision when determining whether to adopt a powerful AI acceleration screening tool for predicting useful polypeptide sequences, but which may also confer weaker IP rights.
Small molecule drugs carry less IP risk when using AI tools in screening. A small organic molecule is typically three-dimensional in its overall structure (instead of a linear sequence). Courts often recognize that this unpredictability requires human input from synthetic organic chemists to determine how an AI-generated formula can be converted into a physical form. This human contribution satisfies the requirement for conception under the legal doctrine of simultaneous conception and reduction to practice.
However, if companies were to replace this human input with another AI model, the risk of losing IP rights would likely increase. The complexity and variability inherent in small molecule synthesis provide somewhat of a safeguard against AI being recognized as the sole inventor.
The legal framework surrounding AI inventorship and the practical aspects of drug synthesis creates a risk-reward dilemma for pharmaceutical companies. The ease of synthesizing certain compounds, particularly antibodies and polypeptides, increases the likelihood that AI could be considered the sole inventor, thereby threatening patent rights.
In contrast, the inherent complexity and unpredictability of synthesizing small molecule drugs provide a buffer against the risk of losing IP rights to AI inventorship. Pharmaceutical companies must navigate these challenges carefully, balancing the benefits of AI-driven efficiency against the potential legal and IP risks.
For more insights at the intersection of AI drug discovery and intellectual property, see our publication “Emerging Legal Terrain: IP Risks from AI's Role in Drug Discovery,” published by IP Strategist.