The OpenAI Foundation is spending $500,000 to acquire datasets from failed biotechnology companies through bankruptcy proceedings, part of a broader initiative to address what experts call the largest obstacle to using artificial intelligence in medicine. The foundation announced today a program called Public Data for Health, which seeks to generate high-quality scientific datasets that can train AI systems to accelerate drug discovery and approval. The effort reflects growing recognition that AI models require far more biological information than currently exists to achieve meaningful breakthroughs in curing disease.
The foundation's first round of grants includes $40 million directed toward collecting information on experimental cancer vaccines at the University of North Carolina, Chapel Hill, and additional funding for OpenAdmet, which organizes competitions where scientists attempt to forecast how drugs will behave in the body. The bankruptcy-focused project received the smallest allocation at half a million dollars and will be executed by 1Day Sooner, an advocacy organization for clinical trial participants. That group currently holds three datasets, two donated by Lumen Bioscience, a company that previously used Chapter 11 proceedings to access another firm's drug development records. The organization estimates it can obtain non-exclusive copies of company datasets for only tens of thousands of dollars each, though two bidding attempts this year were unsuccessful.
Morgan Levine, formerly a vice president for computation at longevity firm Altos Labs, told the publication that "data is the biggest bottleneck in successfully applying AI to biology." According to the OpenAI Foundation, future advances in preventing and curing illness will emerge from combining the intelligence of new models with more observations of the world. Ruxandra Teslo, the policy analyst whose proposal sparked the bankruptcy initiative, explained that roughly 70% of drug development spending and time goes toward clinical development, yet the process remains opaque particularly for smaller biotechnology firms generating innovations.
The foundation's pharmaceutical data collection targets so-called common technical documents, which contain exchanges between companies and regulators plus comprehensive scientific and medical measurements that represent everything known about a drug. Making these files available could transform AI into a regulatory specialist, potentially streamlining how treatments reach the market, according to Teslo. The OpenAI Foundation now stands to become the wealthiest charitable entity globally, holding a 26% equity stake in OpenAI that could be worth $250 billion if the company's planned initial public offering achieves its projected $1 trillion valuation—exceeding the roughly $180 billion held by the Gates Foundation and associated trust at the end of 2025. The foundation, still filling key positions, began scaling up its grants only this year, with its largest single gift of $100 million awarded in August to the Common Health Coalition for hepatitis C drug access.
Jacob Trefethen, a foundation executive, said the organization operates essentially separately from OpenAI but shares the mission of ensuring artificial intelligence benefits all of humanity, and that it hopes to distribute $1 billion by year's end. The data acquisition push comes as concerns about AI-generated bioweapons have intensified, with some industry insiders estimating a 10% or greater probability of human extinction within the next decade, and leaders including Sam Altman and Elon Musk endorsing calls to slow AI capability improvements so safety measures can catch up. If the foundation's bankruptcy strategy proves effective at scale, medical AI could gain access to a trove of failed-experiment insights that have historically remained locked within corporate vaults, transforming costly setbacks into collective learning opportunities. The initiative tests whether philanthropy funded by AI profits can solve the data scarcity problem that currently limits AI's own usefulness—a circular challenge that will require both deep pockets and creative sourcing to overcome.