The OpenAI Foundation will fund a "biotech's lost archive" proposal with a $500,000 grant, folding the project into a broader initiative it calls Public Data for Health to build higher-quality biological datasets for AI training.
The money will pay 1Day Sooner, a non-profit, to bid for and assemble regulatory submissions from failed biotech companies. These filings, known as common technical documents, are the dossiers firms lodge with regulators to seek drug approval.
The foundation says the push complements other grants. These include $40 million to a University of North Carolina programme studying novel cancer vaccines, and support for prediction competitions run by OpenAdmet.
"We expect many remaining breakthroughs in preventing and curing disease to come from pairing the intelligence of new models with more observations of the world, in other words, more data," the OpenAI Foundation said in a statement.
A deeper-pocketed strategy
The grant sits within a larger endowment strategy. The foundation holds a 26% equity stake in OpenAI, and executives say it hopes to give away $1 billion by the end of the year.
The move signals a willingness to buy hard biomedical data rather than rely only on partnerships.
Proving it works
The immediate milestone is a proof of concept. The $500,000 will help the group show it can obtain the data troves, drawn from the proposal by Ruxandra Teslo.
1Day Sooner says it already holds three datasets, which it plans to expand through bankruptcy sales for reuse in model training.
