Artificial intelligence is changing how drugs are developed — though the days of AI-designed treatments for the masses are still probably years off.

Why it matters: AI has driven a surge of research and development by identifying disease targets, improving clinical trials and crunching amounts of data in a way that humans never could.

• But it's presenting new regulatory challenges as some of the compounds go through pre-clinical testing.

What they're saying: The possible number of molecular structures that can be created to develop a drug is "more than all the stars in the universe," said Diogo Rau, chief information and digital officer at Eli Lilly.

• "So obviously, [it's a] great opportunity for AI to just go explore all of these spaces."
• In January, Eli Lilly announced a "co-innovation lab" with Nvidia. That came after the drug giant in October unveiled what it called the pharma industry's most powerful AI supercomputer.

Driving the news: AI is able to evaluate millions of molecules to see which ones are likeliest to work against a particular target.

• For example, MIT researchers in August reported they had identified new antibiotics to use against drug-resistant bacteria after using AI to screen 36 million possible compounds.

How it works: AI's computing power can shorten development times, lower costs and predict chemical properties that humans don't have time to examine, backers say.

• "The beauty of it is, you find molecules that are not things that you and I would have envisioned," said Najat Khan, CEO of the AI-focused pharmaceutical company Recursion.
• "It goes from this artisanal, 'Let me do one thing and try again,' to a much more systematic approach."

Recursion reported positive data in December from a trial of its drug candidate to treat a rare genetic condition that can lead to colon cancer, which the company called the "first clinical validation" of its AI platform.

Between the lines: The FDA is encouraging more of this work while adapting its own regulatory processes to the new world.

• "FDA has been very supportive in public statements over the last couple of years," said Joe Franklin, chief legal and policy officer at the Biotechnology Innovation Organization.
• There are areas where additional policy guidance could be helpful to industry, he said.
• "AI is developing so quickly that ongoing policy development by FDA will be important, including in areas like AI in clinical research and drug manufacturing," Franklin said.

The FDA laid out guidance on things to consider for AI in drug development last year.

• These included transparency in new drug applications about the AI models used developing the product.
• "AI is used by drug developers and we want that," FDA commissioner Marty Makary said at a PhRMA event last month. "We want to see more of that."
• The FDA last year also announced steps to phase out animal testing for certain drugs and pivot to AI-based models instead.

What's next: The yet-to-be-realized milestone is FDA approval of a drug whose development heavily relied on AI. Several such therapies are in clinical trials.

The bottom line: AI may be great at streamlining R&D, but its ability to predict variables like why drugs fail is still a work in progress. Despite that, backers say the technology's potential is becoming more and more apparent.

• "There's going to be some skeptics, but they will come around once they see data," Khan said.

One obstacle for drug developers using AI is ensuring that they can still patent their discoveries.

Why it matters: AI cannot be listed as the inventor on a patent.

The big picture: If a drug was entirely designed by AI, it likely wouldn't be eligible for intellectual property protection and market exclusivity.

• But if humans make significant contributions alongside AI, the product can still be patented.

What they're saying: "Maybe sometime in five or 10 years, somebody will be coming out with an AI-generated molecule," Eli Lilly's Rau said. "But I can guarantee you that they won't call it an AI-generated molecule, because that will immediately invalidate their patent."

Companies that use AI have to make sure that human researchers contribute to the claimed invention, per guidance from the law firm Ropes & Gray.

• That means they "contribute significantly" to every claim set forth in the patent application and don't just synthesize a molecule AI suggests.
• "Even the AI-generated molecules that we've come up with that are doing really, really well, the humans have still found ways to make some improvements to it," Rau said. "And therefore, you're still able to protect your IP."

The bottom line: As AI becomes more commonplace in drug development, lawyers will have their hands full sorting out rules of the road.