OpenAI Foundation awards $500,000 grant for biotech bankruptcy data archive

The Public Data for Health program issued initial grants totaling over $40.5 million. Funds support cancer vaccine data collection at UNC Chapel Hill and OpenAdmet drug prediction competitions. Ruxandra Teslo's proposal targets bankruptcy proceedings to recover manufacturing records and trial data previously treated as trade secrets.

Benchmarks from protein structure and ADMET models show performance plateaus when training sets remain below 10^7 high-quality labeled observations. Public Data for Health addresses this gap by converting private regulatory archives into machine-readable formats. Morgan Levine previously identified data volume as the primary constraint on biological AI at Altos Labs.

The foundation holds 26 percent equity in OpenAI, positioning it for potential $250 billion valuation. Separation from the for-profit entity allows independent grantmaking, yet both entities share the charter goal of broad benefit. Earlier $100 million commitment to hepatitis C access programs demonstrates the shift from model development to external data infrastructure.

Next disbursements target additional failed-company archives and standardized assay repositories. Execution depends on legal access to sealed dockets and conversion pipelines that preserve regulatory provenance.