AI labs are buying data from dead startups to train their models
Startups aren't just dead—their data is being harvested. AI labs are buying up digital archives from defunct companies, paying $25K-plus per tranche, and nobody knows what's in them.
Why it matters
A new data acquisition channel has opened: AI labs are systematically purchasing datasets from dead startups and their founders, creating an unregulated secondary market for training data. The practice raises questions about data provenance, consent, and what ends up in frontier models—and whether founders fully understand what they're selling.
The key facts
11 to knowKeelan Miskell received ~$25,000 for his defunct crypto startup's code base and user data
A second deal from the same startup's archives is pending (amount undisclosed, 'likely around' a similar figure)
The practice is becoming systematic: AI labs are actively acquiring data from defunct startups
Users in the acquired datasets are typically anonymous; provenance tracking is opaque
No regulatory framework or industry standard governs data sales from defunct companies
The article identifies this as an emerging practice, not an isolated transaction
Keelan Miskell netted ~$25,000 for a crypto startup's codebase; additional pending payment (amount undisclosed)
Transaction involves code repository and data from a multi-year-old, anonymous-user startup
Trend: AI labs systematically acquiring dead-startup data for model training
No disclosure of data-provenance auditing, user consent recovery, or IP indemnification mechanisms
Published: Oct 13, 2026 (Technology Review)
Go to the source
MIT Technology Reviewtechnologyreview.com
Publisher excerpt: When Keelan Miskell began winding down his crypto startup this year, he didn’t think its data was worth much: the users were anonymous, the company just a few years old. But he’s since turned the firm’s digital archives into real money. He’s netted roughly $25,000 for his code base and a…