AI labs are buying data from dead startups to train their models

AI labs are buying data from defunct startups to train their models, MIT Technology Review reports. For companies that have already shut down, digital archives once treated as scrap now have a new route to value.

The report offers one concrete case: when founder Keelan Miskell began winding down his crypto startup this year, he did not think its data was worth much, since its users were anonymous and the company was only a few years old.

Instead, he turned the firm's digital archives into real money, netting roughly $25,000 for the code base alone, plus a still-pending sum.

Deals like this show that the scramble for training data is reaching beyond the public web and into what might be called corporate leftovers. For buyers, these archives are relatively clean and easier to license; for sellers, they are extra income on top of a wind-down.

New problems come with them, though: the anonymous users inside the archives, the original licensing terms, and the privacy boundary may all be revisited once the data changes hands.

The question to watch is whether a market for second-hand data keeps growing as public data gets tighter, and what standards AI labs will use to pick through it.