OpenAI Funds New Way to Speed Up Medical Breakthroughs

What if the key to curing diseases faster is hiding in the files of failed biotech companies? OpenAI's nonprofit foundation is betting $40 million that it is.

The OpenAI Foundation just launched Public Data for Health, a new program that pays to create scientific datasets to train medical AI systems. The goal is simple: AI can't make breakthroughs in medicine without much more information to learn from.

One of the most creative projects getting funded is a biotech rescue mission. Policy analyst Ruxandra Teslo proposed buying data from bankrupt biotech companies at their bankruptcy auctions. These files contain detailed safety data, manufacturing strategies, and regulatory conversations that are usually locked away as trade secrets.

Her organization, 1Day Sooner, received $500,000 to prove the concept works. They're already in possession of three company datasets and estimate they can acquire each treasure trove for just tens of thousands of dollars.

"Everyone is recognizing that data is the biggest bottleneck in successfully applying AI to biology," says Morgan Levine, a former vice president at longevity company Altos Labs. Without these training materials, AI systems simply can't become the powerful medical copilots researchers need.

The foundation is also giving $40 million to collect data about novel cancer vaccines at the University of North Carolina. Another grant supports competitions where researchers predict drug effects, building databases that help AI understand how medicines work in the body.

These documents, called common technical documents, contain everything known about a drug. They include back-and-forth conversations between companies and regulators, plus detailed scientific measurements from years of research.

Why This Inspires

This initiative tackles a frustrating reality in drug development. About 70% of the money and time spent bringing new medicines to patients happens during clinical trials and regulatory approval, yet this process remains a black box for most researchers.

Teaching AI systems to understand this complex landscape could dramatically speed up how quickly life-saving treatments reach patients. Instead of reinventing the wheel, scientists could learn from both the successes and failures documented in these rescued datasets.

The foundation plans to give away $1 billion by the end of this year. With a potential $250 billion stake in OpenAI's future value, it's positioned to become one of the world's richest charitable organizations.

Josh Morrison, president of 1Day Sooner, believes these rescued datasets could transform medical research. "People say we will invent AI and AI will cure cancer, but that's very removed from the messy reality and the regulatory process," Teslo explains.

Every failed biotech company represents years of hard-won scientific knowledge. Now, instead of disappearing into bankruptcy proceedings, that research could teach the next generation of medical AI and help bring hope to patients waiting for cures.