SimpleClosure's Asset Hub sells dead startups' data to AI labs

The market exists because AI training labs hit the limits of public and synthetic datasets last year. Real archives avoid common errors in synthetic worlds, such as references to dates that don't exist, though the extent of performance gains remains unclear and no company has yet released tests comparing real-world data training with the alternatives. Pricing reflects that uncertainty: buyers quote wildly different numbers because valuations rest on "research theses" estimating how many useful tasks come from, say, 100,000 lines of code.

Legally, the practice could get iffy. Spirit Airlines' flight attendants' union objected to that sale on privacy grounds, and attorney Greg Ewing has called such data "legally radioactive," pointing to private details like doctor appointments and school pickup arrangements that lurk in emails and calendars. Suppliers employ data scientists to scrub personally identifiable information and exclude HR records, and buyers say they scrub too, but Ewing describes a "Wild West of AI training" only starting to give way to proper licensing agreements and vetted use.

There is also the question of who profits. Proceeds typically flow to senior creditors and investors, not employees, which law professor Ifeoma Ajunwa calls a form of "captured capital." She argues employers should have express consent before selling workers' data, while adding that no federal law bans such sales. Founders, for their part, have little to no visibility into who is buying the data or what they plan to do with it, and little assurance they're getting a fair offer. Data prices will likely stabilize in the coming year, though, unless an AI bubble pops first.