Policy analyst Ruxandra Teslo, who works on clinical trials, proposed purchasing regulatory filings, manufacturing plans and safety data from the bankruptcy proceedings of failed biotechnology companies and using them to train AI. She argued that the archive, which she called “Biotech’s lost archive,” could help develop AI-powered systems for drug approval processes.
OpenAI’s nonprofit parent organization, OpenAI Foundation, announced that it would support the idea through its Public Data for Health program. The program funds the creation of high-quality scientific datasets to advance AI in medicine. Under the first grants:
- University of North Carolina, Chapel Hill will receive $40 million to collect data on new cancer vaccines.
- OpenAdmet, which organizes competitions to predict drug effects, will be supported.
- $500,000 will be allocated to Teslo’s archive project, which will be carried out by 1Day Sooner, where Teslo serves as a consultant.
OpenAI Foundation said that advances in preventing and treating diseases would come from combining more data with the intelligence of new models.
The for-profit company formed through OpenAI’s restructuring is expected to reach a valuation of $1 trillion in the planned IPO. Because of its 26% stake in OpenAI, the foundation could potentially hold shares worth $250 billion; Gates Foundation and an affiliated trust had approximately $180 billion in assets at the end of 2025.
The San Francisco-based foundation accelerated its grantmaking efforts this year, giving $100 million to Common Health Coalition in August. These initiatives are being carried out as fears continue that out-of-control AI could destroy humanity. Altman and Elon Musk supported Dario Amodei’s call to slow the pace of AI development.
Why it matters
This initiative shows that artificial intelligence research in medicine is focusing not only on developing new models but also on making scientific and commercial data that remained fragmented in the past usable. Applications, manufacturing plans, and safety records obtained from failed biotechnology companies could create a previously inaccessible resource for systems that assess drug efficacy and approval processes. The distribution of support across cancer vaccines, drug efficacy prediction, and the lost biotechnology archive reveals that the foundation regards its data infrastructure as one of the fundamental components of healthcare AI. However, it remains unclear to what extent this material will make data from different companies and processes comparable, and how the systems to be developed will be used in drug approvals.
Background
OpenAI is not a new name in the FikirPilot archive: we have published 45 articles mentioning this name in the last 90 days; the latest is dated September 19, 2026.