What If the Cure Is Already Sitting on the Pharmacy Shelf?

Imagine that a medicine already sitting on a pharmacy shelf could help treat a completely different disease.

The drug has already been manufactured.

It may already have years of safety information behind it.

Doctors already know how to prescribe it.

But nobody has figured out that it might help someone with another condition.

That isn’t science fiction.

It’s called drug repurposing, and a nonprofit called Every Cure is trying to make the search for these hidden possibilities much more systematic.

The basic idea is surprisingly simple

Most medicines are developed with a particular disease or biological target in mind.

But medicines can affect more than one pathway in the body.

And different diseases can sometimes involve some of the same biological mechanisms.

So a drug developed for one condition might potentially help with another.

The problem is finding those connections.

There are thousands of approved medicines and thousands of diseases.

Trying every possible combination experimentally would be enormously expensive and time-consuming.

That’s where AI comes in.

Let the computer search the possibilities

Every Cure has developed an AI-powered platform called MATRIX, short for ML/AI-enabled Therapeutic Repurposing In eXtended uses.

The goal is to systematically examine the relationships between existing medicines and diseases and identify combinations that appear worth investigating.

ARPA-H’s MATRIX program describes an “all drugs vs. all diseases” approach designed to predict which existing medicines might have therapeutic potential for diseases with inadequate treatment options.

The idea isn’t to ask AI:

“What is the cure for this disease?”

It’s to ask a much larger question:

“Which existing medicines might be worth investigating for this disease?”

That’s a subtle difference—but an important one.

AI doesn’t get the final say

This is perhaps the most important part of the story.

An AI prediction isn’t a treatment.

Every Cure says its predictions are reviewed by medical experts, and promising opportunities move through laboratory research, preclinical studies and clinical research.

The organization is also developing AI agents to help identify promising drug-disease matches.

In other words:

AI searches. Humans investigate. Evidence decides.

That’s a much more realistic vision of medical AI than the idea that a computer can simply announce a cure.

There’s already a remarkable human story behind it

Every Cure was founded by Dr. David Fajgenbaum, who developed a rare and potentially fatal disease called Castleman disease while he was a medical student.

He became critically ill and nearly died multiple times.

After conventional treatment options were running out, Fajgenbaum and his colleagues began looking for an existing drug that might target the biological pathway involved in his disease.

They identified sirolimus, a drug that had been used for decades as an immunosuppressant in transplant medicine.

It had not originally been developed to treat Castleman disease.

Fajgenbaum ultimately used the repurposed drug and has remained in remission for more than a decade, according to Every Cure.

That experience helped inspire the creation of Every Cure.

The organization was founded in 2022 with the goal of turning this kind of discovery from something that occasionally happens into a systematic process.

And there’s already evidence that the approach can lead somewhere

One of Every Cure’s more advanced examples involves Rosai-Dorfman disease, a rare inflammatory disorder.

Researchers studied 11 patients who received a combination of two existing drugs: lenalidomide and dexamethasone.

Ten of the 11 patients experienced clinical or radiographic improvement, according to the published study.

Every Cure’s computational analysis of nearly 3,000 FDA-approved drugs also identified both medicines among the top candidates for Rosai-Dorfman disease.

That’s an interesting convergence:

The computer identified a promising possibility, and clinical evidence provided support for it.

Then something important happened.

In 2026, the National Comprehensive Cancer Network updated its treatment guidelines to make lenalidomide/dexamethasone a preferred treatment option for Rosai-Dorfman disease regardless of mutation status.

That’s a long way from an AI prediction.

It’s an example of a repurposing idea becoming part of the clinical conversation.

The government is betting on the idea too

This isn’t a tiny research project.

In February 2026, Every Cure announced that it had been selected for up to $76 million in ARPA-H funding for a new three-year phase of work.

The funding is intended to support preclinical and clinical validation of AI-identified drug-repurposing opportunities and move the most promising candidates toward patients.

ARPA-H says the program is designed to advance at least 30 top repurposing opportunities toward preclinical and clinical validation.

By September 2026, Every Cure reported that it had 15 active drug-repurposing programs at different stages of research and development.

Some are very early.

Others are considerably further along.

That distinction matters.

Finding a possibility isn’t the same as finding a cure

This is where the story needs a dose of reality.

Suppose AI identifies an existing drug as a promising candidate for a disease.

That doesn’t mean the drug works.

It could fail in laboratory testing.

It could work in cells but not in humans.

It could produce an effect that isn’t large enough to matter.

Or its side effects could outweigh its benefits.

Even a promising early clinical result needs further evidence.

That’s why the path from:

AI prediction → laboratory evidence → clinical research → treatment

can still take years.

Every Cure itself emphasizes that identifying a promising repurposed drug and proving that it works are separate challenges.

Why repurposing could be so interesting

There is another advantage.

Developing an entirely new medicine can require enormous amounts of time and money.

A repurposed drug may already have substantial information about its safety, manufacturing and pharmacology.

That doesn’t eliminate the need for clinical research.

But it can potentially remove some of the earliest hurdles.

And perhaps more importantly, drug repurposing could make sense for diseases that don’t attract enough commercial investment to justify developing an entirely new medicine.

That’s especially relevant for rare diseases.

There may be relatively few patients, but those patients still desperately need effective treatments.

We’re entering an interesting new phase of medicine

Every Cure isn’t claiming that AI will discover cures for everything.

And it shouldn’t.

What it’s attempting is more subtle—and arguably more useful.

Use computers to search a medical possibility space that humans could never realistically examine one combination at a time.

Then take the most interesting possibilities and put them through the scientific process every potential treatment has to face.

That could eventually uncover treatments that have been hiding in plain sight.

And perhaps that’s the most interesting question of all:

How many useful medicines are already sitting on pharmacy shelves, waiting for someone to discover what else they can do?

We don’t know the answer yet.

But for the first time, we may have the tools to search for it at a scale that wasn’t previously possible.

The takeaway

AI isn’t going to turn an existing drug into a cure simply by finding a connection.

The discovery is only the beginning.

But if AI can help researchers find promising drug-disease combinations that humans would otherwise overlook—and those possibilities can then be tested properly—we may discover that some future treatments aren’t new medicines at all.

They may be old medicines with a new purpose.