Can AI-Discovered Drugs Be Patented?

Transcript

The following transcript of this discussion was edited for clarity.

Artificial intelligence (AI) is drastically speeding up drug development. But companies may face challenges getting patents for AI-assisted breakthroughs, which could weaken incentives to innovate.

I’m Sarah Cambon, and today I’m joined by Marty Gomez, a partner at Goodwin, to unpack his recent article on AI and drug development, which is included in Goodwin’s Forces of Law series.

Sarah Cambon: Marty, thanks so much for being here.

Marty Gomez: Thanks for having me.

Let’s start with the basics. To get a patent, a company has to name the inventor. Can AI be named as the inventor?

This is one of the only areas where we have very clear guidance as it relates to inventorship when it comes to AI. The courts have said there’s really no dispute over this: AI cannot be listed as an inventor on a patent application.

So where does it start to get gray in practice?

These days, humans frequently interact with AI in the context of drug discovery, often involving the creation of a molecular structure for a drug. Depending on the nature of that interaction, the human may — or may not — meet the definition for having conceived the invention. That’s an area where there’s a lot of grayness in the current legal landscape.

Maybe we could think through a couple of hypothetical examples to provide a bit more clarity. If I’m a scientist and I tell an AI model, “Make this drug,” that’s not nearly enough to qualify for inventorship, right?

Probably not. Based on our reading of the rules, it seems unlikely that if all you do is give the end point of what you want the AI to achieve, it is probably not enough to meet the criteria of having conceived the invention.

But if I’m a scientist, and I’m iterating with AI, and we ultimately arrive at a new drug, that’s where it’s unclear whether I’ve done enough to qualify for inventorship?

That’s right.

But if a human is setting the parameters of the problem, if they are identifying data, and if they are giving constraints, they’ll certainly get to a point where they are interacting so much with the AI that the AI starts to take on the role of a software or technology tool, which innovators have been using routinely for decades or maybe even longer.

In that case, the human’s interaction would be considered conception, and, therefore, it would be appropriate to list them as an inventor on the patent application.

What should companies be doing right now to protect their ability to patent AI-assisted inventions down the line?

To support the fact that the human actually qualifies as an inventor and meets this conception standard, companies probably want to have some sort of evidence of how the human interacted with the AI. That includes documentation that supports what the human’s role was — for example, scoping the problem, providing the relevant datasets and constraints, and iterating with the model.

That kind of evidence will be helpful if and when a challenge arises claiming that the human should not qualify as an inventor due to the involvement of AI.

It seems like that’s a fine line — to clearly describe the role the AI played and the contributions of the human or scientist while being careful not to disclose your secret sauce.

Absolutely, though companies have always faced a decision around how much information to include in their patent applications. The use of AI just puts the issue under the microscope.

What’s at stake here if companies can’t secure patents on AI-assisted inventions?

A patent grants its owner the exclusive right to make, use, sell, and profit from the patented subject matter. The theory behind this system is that these exclusive rights incentivize individuals and companies to invest in the significant effort required to develop new innovations.

Drug discovery, for example, often requires enormous effort and financial investment. Without patent protection, others could freely make or profit from the discovery, undermining the incentive for companies or individuals to undertake that level of investment in the first place.

What should companies be watching in the next year or two that might clarify the rules?

We are likely to get some case law in the coming years that will give us some additional guidance around what factual circumstances resulted in an individual being considered an inventor or not based on interactions with AI.

In the nearer term, we may receive some decisions from district courts, which are regional and only have precedential effect in certain areas of the country.

It may take even longer for us to get decisions from either the U.S. Court of Appeals for the Federal Circuit or the U.S. Supreme Court.

I’d keep an eye on that, but it may still be a couple of years until we get definitive guidance, or better guidance, on these issues.

Marty, thank you so much for your time.

Thank you so much for having me.