How to Maximize Your AI-Assisted Invention’s Patentability

Artificial intelligence is rapidly reinventing the research and development process, but until recently, patentability remained a major uncertainty. That’s until the United States Patent and Trademark Office stepped in and issued guidance in February outlining what it takes to obtain patent protection when AI is involved in the invention process.

In a recent webinar, intellectual property partners Bob Hulse, Stuart Meyer, and associate Michael Saffron broke down what it all means for inventors—and shared key strategies for maximizing innovation and IP protection. Here are the highlights:

AI-assisted inventions can be patentable. In February, the USPTO clarified that AI-assisted inventions “are not categorically unpatentable,” and that its analysis would focus on human contribution. Specifically, a human must “significantly contribute” to the conception of the invention—the “lightbulb moment." That’s a highly fact-dependent question and will depend on how the human used an AI system, the technical field of the invention, and the state of the art.

Focus on prompt construction. A significant contribution to inventorship requires more than merely identifying a problem and asking AI to solve it—and it can all start with your prompt. Try to avoid prompting an AI system to come up with a solution to the problem you’re trying to solve. Instead, be specific and offer the AI guidance on how it might achieve various elements of the solution you’re working toward. Refine and iterate on your ideas. The more interactions you have with AI during the process, the more significant your contributions become.

‘Practice’ makes perfect. Another opportunity to stake out human contribution comes when it’s time to physically realize your creation—what patent lawyers call “reduction to practice.” Simply following an AI’s instructions to build or otherwise realize a concept is not enough. Instead, refine and improve on the AI’s draft. Redesign a specific part or parts so they work better within the larger system you’re crafting.

Training specific models can lead to inventorship. Owning or overseeing an AI system is not enough to claim inventorship of its outputs—and neither is merely implementing a general-purpose AI model’s idea. But training a bespoke model that’s designed to help you create narrowly tailored solutions could be.

Mind your duty to disclose. Anyone involved in filing or processing a patent application has a duty to disclose all information they have related to the pending claim’s patentability. Thus, patent applicants and their attorneys have a duty to inform the USPTO if their invention may not have a human inventor. Additionally, patent attorneys have an additional duty to make a reasonable inquiry into the inventive process if they have reason to suspect that a human may not have significantly contributed to conception. Patent attorneys, and clients, should take steps to avoid these potential minefields by ensuring that they comply with these duties of disclosure and reasonable inquiry.

Stay up-to-date and discuss recommended practices with your attorneys. Across legal disciplines (patents, copyrights, trademarks, privacy, etc.), lawmakers and stakeholders are constantly releasing more guidance on how the law will be applied to AI. It is recommended that clients try to stay on top of these changes to ensure they’re complying with recommended practices.

For more insights, check out the full webinar (CLE credit available).