TL;DR
Companies helping startups close are increasingly arranging sales of their code and workplace data to AI labs and other buyers. The market may give failing firms a way to recover value, but independent evidence of training benefits is lacking, and privacy, consent and who receives the proceeds remain concerns.
Companies that help startups wind down are increasingly selling their code and workplace archives to AI labs and other technology buyers, creating a new source of AI training data from businesses that have shut down. SimpleClosure, which added data sales to its services in April, says 47% of its deals now include code or data; the value of that material for AI performance, and the safeguards around its use, remain uncertain.
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Keelan Miskell, who began winding down his crypto startup this year, told MIT Technology Review that he received about $25,000 for its code base. A separate sale of data from the company’s Google, Notion, Figma and Slack workspaces is pending; Miskell estimated it could bring roughly $15,000. The sales were arranged through SimpleClosure’s Asset Hub, which connects closing companies with buyers.
SimpleClosure CEO Dori Yona said the platform has handled more than 150 deals and expects to return over $10 million to startups by the end of 2026. Buyers include foundation AI labs, companies building simulated workspaces for AI training, and established technology firms developing their own models. Sunset, another company that helps businesses close, is also expanding its data resale operation.
AI data company Surge AI buys archives to build training environments in which AI agents practice tasks, according to Nick Heiner, who leads its work on those environments. He said his team looks for records with varied situations and difficult judgment calls. A rich Slack history can be more useful than a larger bundle of email, calendar and chat records, he said, depending on the conversations it contains.
A New Revenue Stream With Trade-offs
For a company shutting down, selling digital assets can bring in money that might otherwise be lost. Miskell’s reported sale illustrates that code and years of internal work may have value beyond the company’s original business. But the financial upside does not necessarily reach the people whose activity produced the records: proceeds typically go to senior creditors and investors, rather than employees.
The market also creates a new route for AI developers to obtain records of real work. Buyers and intermediaries say these archives can provide varied examples of collaboration, decisions and technical tasks. However, there are no publicly released tests in the source report comparing model performance trained on these archives with performance using public or synthetic data. The proposed training advantage is not yet established.
Workplace archives may also contain sensitive details about workers and people they communicate with. Selling them after a business closes raises questions about privacy, consent and control that do not disappear simply because the company has ceased operating. The market’s growth makes those questions more consequential for former employees, creditors, buyers and regulators.
Products Worth Considering
Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Bose QuietComfort Wireless Over-Ear Noise-Cancelling Headphones, Sandstone
As an affiliate, we earn on qualifying purchases.
How Startup Archives Became Assets
The resale market has emerged alongside a broader industry serving companies that are closing, including firms that handle legal, tax and operational tasks. Carta and AngelList estimates cited by MIT Technology Review put annual increases in startup shutdowns at between 26% and 56% in recent years. The source does not specify a single shared time window or methodology for that range.
SimpleClosure added data sales in April. Its Asset Hub offers both code and workplace records such as emails, chats and documents. Sunset founder and CEO Brendan Mahony said its valuations consider a company’s head count, years in operation, data volume and how workers used the tools. He compared operational records to wood shavings that were once treated as waste but can prove useful to buyers.
The report describes demand as emerging suddenly last year, as AI training labs faced limits in public and synthetic datasets. A larger transaction followed in August: during Spirit Airlines’ bankruptcy, Google outbid Mercor for $10 million to acquire decades of company data, including emails, Teams messages and employee records. Bloomberg reported in September that SpaceX’s AI unit was considering buying data from struggling startups; the report did not say a purchase had been made.
“At least all that stuff we were doing for two and a half years is worth something to someone somewhere.”
— Keelan Miskell, founder of a crypto startup that is winding down
Performance, Privacy and Proceeds
The extent to which real company archives improve AI systems is not yet demonstrated publicly. The report says no company has released comparative tests against alternative training data. Surge AI’s stated interest in varied, realistic records reflects a buyer’s rationale, not proof of measurable gains. Prices are also unsettled: Heiner said buyers offer widely different amounts based on research assumptions about how many useful tasks an archive might contain.
Privacy protections are another open issue. Data sellers and buyers say they use specialists or screening processes to remove personally identifiable information and exclude human-resources records. But the source report does not establish how consistently those practices are applied or whether they remove sensitive details embedded in ordinary email and calendar entries. During Spirit’s bankruptcy, the flight attendants’ union objected to the data sale on privacy grounds.
It is also unclear how often workers have given informed permission for their records to be resold, how former employees can challenge a sale, or what happens to data after a buyer acquires it. Attorney Greg Ewing said there is no federal law banning these sales, while Ajunwa argued employees should give express consent. Founders, meanwhile, may have limited visibility into buyers’ identities, intended uses and whether an offer reflects fair value.
What Could Shape the Market
SimpleClosure expects to return more than $10 million to startups by the end of 2026, according to Yona. The report says data prices may become more stable over the coming year, although that outlook is not a settled forecast and could change if demand for AI training data weakens. More deals would provide evidence of market activity, but not by themselves establish that the data improves models.
The next developments to watch are whether AI buyers publish comparative results, whether data-sale agreements specify permitted uses and retention, and whether sellers adopt clearer processes for worker privacy and consent. These details will help show whether the market can offer closing businesses meaningful recoveries while limiting the risks created by transferring workplace records.
Key Questions
What data are AI companies buying from closed startups?
Reported sales include software code and workplace archives such as emails, chat messages, documents and other records from tools including Slack, Google, Notion and Figma. The exact material varies by company and deal.
Has this data been shown to improve AI models?
Not through publicly released comparative tests cited in the report. Buyers say realistic records may help create useful training tasks, but the performance gains remain unclear.
Who receives money from these sales?
The report says proceeds typically go to senior creditors and investors, not employees. How a particular sale distributes proceeds depends on the company’s circumstances and obligations.
What are the main privacy concerns?
Emails and calendars can contain sensitive personal details, even when formal HR records are excluded. The report also raises questions about employee consent, screening practices and buyers’ later use of acquired data.
Are the sales legally prohibited?
The report does not identify a federal law that categorically bans these sales. Legal risks and worker protections may depend on the data, agreements and applicable law; the source does not settle how individual transactions should be treated.
Source: rss
