From MIT Technology Review · · 2m
AI labs are buying data from dead startups
As startups shut down, AI companies are buying their archived emails, code and workspace data to train models. The market is booming, but raises questions about worker privacy and fair pricing.
In brief
As startups shut down, AI companies are buying their archived emails, code and workspace data to train models. The market is booming, but raises questions about worker privacy and fair pricing. AI labs are buying data from defunct startups to train models, creating a new market where founders sell company archives for…
A startup founder sells his data
Keelan Miskell wound down his crypto startup and sold its digital archives to AI labs. He netted roughly twenty five thousand dollars for his code base and around fifteen thousand dollars for workspace data.
“At least all that stuff we were doing for two and a half years is worth something to someone somewhere.”
A new market for dying startups
SimpleClosure brokers data sales from failing companies. Shutdowns are rising between twenty six percent and fifty six percent annually. An end-of-life ecosystem including coaches and death doulas has emerged to handle the aftermath.
SimpleClosure's data sales boom
SimpleClosure added data sales in April.Its Asset Hub platform helps founders sell code and workspace data to AI labs and tech companies.
CEO sees data sales becoming routine
SimpleClosure CEO Dori Yona says data transactions were once rare and somewhat taboo. Now forty seven percent of SimpleClosure deals include code or data sales. Rival Sunset also has a rapidly growing data resale business.
Why AI labs want this data
Demand for defunct companies' data emerged last year as AI training labs hit limits of public and synthetic datasets.
Real data versus synthetic alternatives
Real archives avoid errors in synthetic worlds like references to dates that do not exist.But no company has released tests comparing training results with real data versus alternatives.
Google buys Spirit Airlines data
In August, Google outbid AI data company Mercor to buy Spirit Airlines' decades of data during bankruptcy.
Privacy concerns emerge
Spirit's flight attendants' union objected to the sale on privacy grounds.Attorney Greg Ewing calls such data legally radioactive.
“legally radioactive”
Workers do not share the profits
Proceeds from data sales flow to senior creditors and investors, not employees.Law professor Ifeoma Ajunwa says this represents captured capital.No federal law bans such sales, though employers should have express consent before selling workers' data.
“Employers should have express consent before selling workers' data.”
The takeaway
AI labs are buying data from defunct startups to train models, creating a new market where founders sell company archives for thousands of dollars, though privacy and fairness concerns remain unresolved.
