Frontier AI labs pay large sums for proprietary data, Sara Hooker says

Sara Hooker says frontier AI labs are paying large sums for old Slack threads, bankrupt airline logs, and deprecated company archives.

In a post on X, she says the labs want this real-world signal to shape how their models interact with the world.

Hooker also says labs are paying massive sums for proprietary data to enter vertical markets such as healthcare, taxes, finance, legal, and transportation. She reports that the labs hope to mimic those data-rich surfaces before their upcoming IPOs so they can compete directly in those sectors.