[Adobe Stock]
Like any pharmaceutical, AI-assisted therapeutics require patent protection. The USPTO and the Federal Circuit have reaffirmed that only natural persons can be listed as inventors on patents. The USPTO has also issued guidance on determining inventorship in AI-assisted inventions, which has significant legal and practical consequences. This article outlines a framework for assessing inventorship in AI-driven drug discovery and explores the implications of current legal standards, particularly under 37 CFR 1.56, which mandates disclosure of improper inventorship.
Inventorship: The legal standard
Under U.S. patent law, an inventor is someone who “conceived the subject matter” claimed in an application. Conception requires a definite and permanent idea of how the invention will be applied in practice. Simply having a research goal or general plan is insufficient.
The USPTO uses the Pannu factors to assess inventorship in AI-assisted inventions. To qualify as an inventor, an individual must:
- Significantly contribute to the conception or reduction to practice of the invention.
- Make a contribution that is not insignificant in quality relative to the invention’s full scope.
- Do more than merely explain known concepts or the state of the art.
The challenge of AI in inventorship
Determining inventorship in AI-assisted discoveries is complex and can impact patent validity and ownership. AI’s role in generating drug candidates raises key questions:
- Who is responsible for the conception of an AI-assisted invention?
- How much human input is necessary to establish inventorship?
- Does training or fine-tuning an AI system qualify as an inventive contribution?
The USPTO has clarified that an AI system itself cannot be an inventor, but humans involved in training or prompting AI systems or modifying AI outputs may qualify. Given this, companies must carefully document human contributions to mitigate future challenges to patent validity.
A framework for determining AI-assisted inventorship
To navigate these complexities, companies should assess inventorship based on three key considerations:
1. AI system type and training: The role of human contribution
The nature of the AI system used—whether off-the-shelf or bespoke—impacts inventorship.
- Off-the-shelf AI: These systems are developed to solve general problems. Simply using such a system without further refinement or training does not constitute an inventive contribution.
- Bespoke AI: Custom AI systems developed for a specific purpose, such as identifying a therapeutic with particular properties, may involve significant human contributions. If a researcher trains an AI model to recognize molecular structures with specific binding affinities or toxicity thresholds, this effort could be considered an inventive step.
Thus, the level of human involvement in training an AI system is crucial in determining whether an individual qualifies as an inventor.
2. Prompt engineering: Does input define inventorship?
AI models require human-generated prompts to produce outputs. The design of these inputs—known as prompt engineering—can impact inventorship.
- A generic prompt (e.g., “Find new drugs for this target”) does not constitute an inventive step.
- A complex, targeted prompt that incorporates multiple design factors (e.g., optimizing for binding affinity, metabolism, solubility, and toxicity) could be considered an inventive contribution.
Companies should document who crafts the prompts and the reasoning behind specific input choices to support claims of inventorship.
3. AI output and human modifications
The raw output of an AI system is not inherently an invention. Instead, the key question is what happens after the AI generates a result.
- If a researcher modifies an AI-generated molecule to improve its efficacy, reduce toxicity, or enhance bioavailability, this work could be considered a significant contribution to conception.
- If the AI simply provides a list of potential compounds and the researcher selects one without making meaningful modifications, this is not an inventive act.
When working with third-party AI providers, companies should ensure agreements distinguish between AI-generated “candidate” compounds and final therapeutic inventions to clarify ownership.
Practical considerations for life science companies
Since patent ownership is tied to inventorship, life sciences companies must adopt clear strategies when working with AI-assisted discoveries. The following best practices can help mitigate inventorship disputes:
Contractual clarity in AI collaborations
- Internal and external agreements: Contracts with third-party AI developers or contract research organizations (CROs) should explicitly define who owns the data, training process, and AI-generated results.
- Defining AI training responsibilities: If company employees contribute to training or refining an AI system, their role should be documented to establish potential inventorship claims.
2. Documentation of human contributions
Maintaining comprehensive records of human involvement in AI-driven discoveries is crucial. Key documentation includes:
- Training logs: Records of modifications and refinements made to an AI system.
- Prompt design records: Documentation of input prompts and their rationale.
- Modification reports: Detailed logs of human interventions made to AI-generated outputs.
This documentation can serve as evidence in legal disputes over patent validity and ownership.
3. Managing USPTO compliance and inventorship disclosure
The USPTO requires applicants to disclose improper inventorship under 37 CFR 1.56. Failure to properly attribute inventorship can result in a patent being invalidated. Given the evolving role of AI, companies should conduct internal audits to confirm whether human contributions meet the inventorship standard before filing a patent application.
Future implications: AI’s expanding role in drug discovery
As AI continues to evolve, inventorship challenges will become more prevalent. Pharmaceutical litigation may increasingly involve disputes over whether a company’s use of AI in drug discovery undermines human inventorship claims. Competitors could challenge patents based on published research describing AI’s role in the discovery process.
To mitigate risks, companies must proactively:
- Structure AI workflow to ensure human contributions at critical stages.
- Clearly document all human contributions.
- Structure agreements to avoid ambiguity in AI-assisted innovation.
Conclusion
Generative AI is reshaping drug development, but patent law remains centered on human inventorship. Companies must carefully navigate AI system training, prompt engineering, and human modifications to AI outputs to ensure human inventors.
 Jon Cousin |
 Jordan Phelan |
Key takeaways:
- AI-assisted drug discovery raises complex legal questions about inventorship.
- Companies must evaluate whether AI use involves significant human contributions at multiple stages.
- Clear documentation of human involvement is essential to securing and defending patents.
As AI becomes more embedded in life sciences R&D, firms must anticipate legal challenges and ensure compliance with evolving patent laws. Proper planning today will help protect valuable IP assets in the future.
About the authors
Jon Cousin: Dr. Jon Cousin is a Partner at Cooley LLP. Jon develops and executes worldwide patent strategies for life sciences and pharmaceutical companies. Working closely with his clients to understand their business and objectives, Jon helps them build robust, business-focused and diligence-ready patent portfolios. His practice includes developing patent strategies for drug-discovery and clinical programs, life cycle extension strategies, aligning patent and regulatory exclusivities, and Orange Book listings.
Jordan Phelan: Dr. Jordan Phelan is an Associate at Cooley LLP. Jordan develops and implements worldwide patent strategies for early- and mid-stage biotechnology companies. She has extensive technical experience in DNA sequencing, chimeric antigen receptor T-cell (CAR-T) therapy, gene editing, gene therapy, complex biologics and antibody therapeutics. Her practice focuses on counseling clients regarding global patent strategy and life cycle management, drafting and prosecution of biotechnology applications, and preparation of competitive landscape and freedom-to-operate analyses. She also provides patent support in partnership, acquisition and licensing negotiations.
Filed Under: Data science, Legal precedents and interpretations
