Product1 publisher3 min readPublished
OpenAI Foundation puts $500,000 behind buying failed biotechs' files out of bankruptcy
The grant to 1Day Sooner is the smallest disclosed line in the Foundation's first health-data round. It pays for a feasibility test, because the group's two attempts to get drug company files this year both failed.
The Product Desk · Product desk
What happened
- The OpenAI Foundation launched a program called Data for Public Health, funding the creation of high-quality scientific datasets on the premise that data is what holds AI back in medicine.
- Its first round of data grants gives $40 million to a University of North Carolina, Chapel Hill program collecting data on novel cancer vaccines, and also supports OpenAdmet's drug-effect prediction competitions.
- A separate $500,000 goes to the advocacy group 1Day Sooner to pursue Ruxandra Teslo's idea of bidding at bankruptcy proceedings for failed biotechs' regulatory filings and safety data.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- constraint At the price Morrison quotes, half a million dollars buys more than a dozen copies. The limit on the corpus is whether a trustee will sell one at all.
- decision The Foundation is paying for both ways of getting biology data at once, 80 dollars of new collection for every dollar of archive retrieval.
- precedent Lumen Bioscience already used Chapter 11 to learn about another company's drug development. A funded nonprofit bidding repeatedly would make a failed biotech's trade secrets an asset outsiders expect to price.
Getting a copy of a failed drug company's file means showing up at a Chapter 11 proceeding and outbidding whoever else wants a piece of the estate. 1Day Sooner has tried. Josh Morrison, the group's president and co-founder, says two attempts to obtain drug company files this year were unsuccessful [10]. Two of the three datasets the group holds now arrived as donations from Lumen Bioscience, a biotech that had itself used the Chapter 11 route to gain insights into another company's drug development work [9].
Morrison thinks non-exclusive copies of company datasets could be acquired for only "a few tens of thousands of dollars" each [8]. Take the low end at $30,000 and the $500,000 grant covers about 16 of them [2]. The round's other disclosed figure is $40 million, going to a University of North Carolina, Chapel Hill program collecting new data on novel cancer vaccines [5]. That is 80 times the archive grant [1].
The program is premised on data being the constraint: "Everyone is recognizing that data is the biggest bottleneck in successfully applying AI to biology," said Morgan Levine, a former vice president for computation at the longevity company Altos Labs [4]. The $500,000 pays for something smaller. Morrison says it will help the group prove it can obtain the data troves of bankrupt companies at all [8].
Ruxandra Teslo, the clinical trial policy analyst who proposed the idea, said the filings, manufacturing strategies and safety data could help train AI systems to act as co-pilots in the often opaque drug approval process [1].
Jacob Trefethen, an executive at the foundation, said it hopes to give away $1 billion by the end of the year [11]. The archive grant is 0.05% of this year's target [3]. "We're starting grantmaking when we think the best way to achieve that mission is to make grants to external non-profits, research institutions, and other third parties," Trefethen said in an interview [12]. Its largest single gift so far is $100 million, awarded in August to the Common Health Coalition [14]. Its 26% stake in OpenAI could be worth $250 billion if the planned IPO values the company at $1 trillion [13].
Last week Sam Altman and Elon Musk both endorsed Anthropic CEO Dario Amodei's call to "slow the pace at which we improve the capabilities of AI models" so that risk prevention can catch up, and the fears behind that call include a deadly bioweapon [15]. Awkwardly, the Foundation holds a quarter of Altman's company, and it is now paying to put more biology in front of models [2][13].
For anyone weighing the same route, the variables are the price of a copy and the odds of getting one. Morrison's price makes the first almost irrelevant at this scale. The money is testing the odds. The count to follow is auctions won against auctions entered, and on the group's own account that is zero for two this year [10].
What to watch
- Whether 1Day Sooner wins a Chapter 11 auction, and what the winning bid reveals about the real price of a non-exclusive copy.
- Whether trustees agree to sell non-exclusive copies separately from an estate's exclusive intellectual property.
- Whether the Foundation names a figure for OpenAdmet and the rest of Data for Public Health as it works toward the $1 billion goal.