Artificial intelligence is facing a severe shortage of training materials in biology, where proprietary clinical and manufacturing records remain locked behind corporate walls. To overcome this barrier, OpenAI Foundation is funding a new initiative called Data for Public Health. A broader strategy targets trade secrets and hidden regulatory dossiers of collapsed companies, turning corporate liquidations into foundational datasets for medical AI.
Policy analyst Ruxandra Teslo proposed acquiring detailed regulatory dossiers, manufacturing strategies, and safety data from bankrupt biotechnology firms to build assistants for the drug approval process. Teslo's idea for a biotech archive received $500,000 and will be implemented by the advocacy group 1Day Sooner, representing clinical trial volunteers, which she advises.
"Everyone acknowledges that data is the biggest bottleneck to the successful application of AI to biology," said Morgan Levin, former vice president of computation at Altos Labs, emphasizing why frontier AI developers are looking beyond publicly available research literature.
Aside from the archive grant, OpenAI Foundation announced it will allocate $40 million to a program collecting data on new cancer vaccines at the University of North Carolina at Chapel Hill, and will support OpenAdmet, an initiative running drug effect prediction competitions.
Financial muscle and bankruptcy acquisitions
Since the foundation holds a 26% stake in OpenAI, its philanthropic balance sheet enables capital to be deployed on a massive scale toward healthcare and scientific discovery. This influx of grants follows earlier healthcare moves: in August, OpenAI Foundation allocated $100 million to the Common Health Coalition to help patients get hepatitis C medications.
On the operational level, access to troubled pharmaceutical companies' data is already being tested through ongoing bankruptcy proceedings.
"We anticipate that many of the remaining breakthroughs in preventing and treating disease will come from combining the intelligence of new models with large observations of the world — in other words, more data," stated OpenAI Foundation regarding the fundamental logic of acquiring unindexed empirical records.
Safety debates and practical limits
These data expansion initiatives arrive as top industry executives debate wider risks of developing biological capabilities.
Turning bankruptcy courts into training data pipelines shows that raw compute has hit a hard wall of proprietary biology. The real economic value in biotech no longer lies solely in successful compounds, but in meticulously documented failures left behind after corporate liquidation.