PickAI labs buy data from failed startups
MIT Technology Review reports that founders of shut-down startups are selling their companies' archives to AI labs as training data.
When a startup shuts down, its code and internal records used to disappear with it. MIT Technology Review reports that a growing market now sells those archives to AI labs as training data.
How the market works
The report describes a founder who wound down his crypto startup this year and earned about $25,000 for the company's codebase, with another sale of roughly $15,000 pending for workspace data from tools such as Google Workspace, Notion, Figma and Slack.
That sale was arranged by SimpleClosure, one of several companies that help startups close down. SimpleClosure added data sales in April through a platform called Asset Hub, which helps founders sell code and workspace data including emails, chats and documents. Buyers include frontier AI labs, vendors that build simulated workplaces for training AI agents, and established tech companies training their own models. SimpleClosure says it has handled more than 150 deals, expects to return over $10 million to startups by the end of 2026, and that 47 percent of its deals now include code or data sales. A rival, Sunset, also runs a fast-growing data resale business.
The demand makes sense: real company communications and codebases are exactly the kind of messy, realistic material labs need to train agents that do office and engineering work.
Why it matters for builders
If you run a startup, your internal data may have resale value, but it also carries obligations. Emails and chats contain personal information about employees, customers and partners who never agreed to have it used to train AI. Contracts with customers may restrict what you can do with their data, even after the company closes. Founders considering a sale should get legal advice and remove or anonymize personal and customer information.
If you are an employee, this is a reason to keep personal matters out of work tools. And if you are a customer of a small startup, check its terms on what happens to your data in an acquisition or shutdown.
For AI builders, the trend highlights how valuable realistic workflow data has become, and how important consent and provenance will be as scrutiny of training data grows.
Source: MIT Technology Review. This article is PickAI’s own summary and analysis based on the original reporting.
