Epstein asked to call during a fundraiser. My notes say that I tried to explain AI alignment principles and difficulty to him (presumably in the same way I always would) and that he did not seem to be getting it very much. Others at MIRI say (I do not remember myself / have not myself checked the records) that Epstein then offered MIRI $300K; which made it worth MIRI’s while to figure out whether Epstein was an actual bad guy versus random witchhunted guy, and ask if there was a reasonable path to accepting his donations causing harm; and the upshot was that MIRI decided not to take donations from him. I think/recall that it did not seem worthwhile to do a whole diligence thing about this Epstein guy before we knew whether he was offering significant funding in the first place, and then he did, and then MIRI people looked further, and then (I am told) MIRI turned him down.
Epstein threw money at quite a lot of scientists and I expect a majority of them did not have a clue. It’s not standard practice among nonprofits to run diligence on donors, and in fact I don’t think it should be. Diligence is costly in executive attention, it is relatively rare that a major donor is using your acceptance of donations to get social cover for an island-based extortion operation, and this kind of scrutiny is more efficiently centralized by having professional law enforcement do it than by distributing it across thousands of nonprofits.
In 2009, MIRI (then SIAI) was a fiscal sponsor for an open-source project (that is, we extended our nonprofit status to the project, so they could accept donations on a tax-exempt basis, having determined ourselves that their purpose was a charitable one related to our mission) and they got $50K from Epstein. Nobody at SIAI noticed the name, and since it wasn’t a donation aimed at SIAI itself, we did not run major-donor relations about it.
This reply has not been approved by MIRI / carefully fact-checked, it is just off the top of my own head.
Here’s Yudkowsky’s memory of what happened: