TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
Companies that help startups shut down are now arranging sales of their code and workplace archives to AI labs and other buyers. The market is growing, but there is no public evidence yet showing how much real-world data improves model performance, and privacy, consent and payment questions remain.
Companies that help startups wind down are selling their code and workplace data—including emails, chats and documents—to AI labs and other buyers, according to an October 13 report by MIT Technology Review. The developing market gives failed companies a way to recover some value from their digital archives, while raising unresolved questions about worker privacy, consent and the usefulness of the data for AI training.
One founder interviewed for the report, Keelan Miskell, said he received about $25,000 for his crypto startup’s code and was waiting on a separate payment, likely around $15,000, for data held in Google, Notion, Figma and Slack. The sale was arranged through SimpleClosure, a company that assists businesses with shutdowns. The amount for the workspace data was not final.
SimpleClosure added data sales to its services in April. Its Asset Hub platform has handled more than 150 deals, and the company expects to return more than $10 million to startups by the end of 2026. CEO Dori Yona told MIT Technology Review that 47% of the company’s deals now include code or data sales. Rival Sunset also sells data from companies it helps close.
Buyers include foundation AI labs, businesses that build simulated workspaces for training AI agents, and established technology companies developing their own models. Surge AI, for example, buys data to create practice environments in which agents can work through tasks such as software debugging or legal drafting. The report says buyers value archives differently: a varied, detailed Slack history may be more useful than a larger collection of less informative records.
A New Source of AI Training Data
These sales could give model developers access to records of real work—conversations, decisions and project histories that may be difficult to reproduce synthetically. For startups closing their doors, selling those records can bring in money during a shutdown. But the transaction also transfers data created by employees and customers, who may not have had a say in its later use.
The market’s growth is not proof that the data improves AI systems. MIT Technology Review reported that no company has released tests comparing model performance after training with real company archives against performance using synthetic or other data. Buyers’ valuations rely partly on estimates of how many useful tasks an archive might support, making prices difficult to judge from outside.
Who receives any proceeds is another concern. The report says sale proceeds typically go to senior creditors and investors, rather than employees whose communications and work records may be included. That makes the market relevant not only to AI developers and founders, but also to workers, customers and creditors.
As an affiliate, we earn on qualifying purchases.
How Startup Archives Became Assets
The resale market is part of a wider set of services for companies shutting down. MIT Technology Review cited estimates from Carta and AngelList that startup shutdowns have risen by 26% to 56% annually in recent years. As more businesses close, firms offering legal, tax and operational help have developed ways to value and sell what remains.
Demand for defunct companies’ data emerged last year, the report said, as AI training labs encountered limits in public and synthetic datasets. In August, Google outbid AI data company Mercor for data from bankrupt Spirit Airlines, including 100 million emails, 500 million Teams messages and 176,000 employee records. The reported price was $10 million. Bloomberg reported in September that SpaceX’s AI unit was weighing purchases of data from struggling startups; that was a reported consideration, not a confirmed transaction.
Real archives may avoid some artificial inconsistencies found in synthetic datasets, such as references to nonexistent dates. However, that potential advantage remains a rationale offered in the market, not a demonstrated performance result. Data packages can also differ sharply in relevance, quality and the kinds of decisions they capture.
“At least all that stuff we were doing for two and a half years is worth something to someone somewhere.”
— Keelan Miskell, former crypto startup founder
enterprise data privacy protection software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Privacy, Pricing and Proof Gaps
Several central questions remain open. The report found no published comparative tests showing whether training with company archives improves AI performance over alternatives. It is also unclear how consistently buyers disclose their intended uses to sellers, or how founders can assess whether an offer reflects an archive’s value.
Privacy protections vary in detail. Suppliers and buyers told MIT Technology Review they use data scientists or other processes to remove personally identifiable information and exclude human-resources records. But attorney Greg Ewing warned that sensitive details can remain embedded in messages and calendars, including medical appointments, family arrangements and personal plans. Spirit Airlines’ flight attendants’ union objected to the airline-data sale on privacy grounds.
The report also describes unresolved legal and consent issues. Ewing called some of the data “legally radioactive,” while Ajunwa said employers should get express consent before selling workers’ data. The report notes that no federal law currently bans such sales; it does not establish how courts or regulators will handle every kind of archive or transaction.
As an affiliate, we earn on qualifying purchases.
The Market Awaits Clearer Rules
SimpleClosure expects to return more than $10 million to startups by the end of 2026, according to the company. Market participants expect prices to become more consistent over the coming year, although the report says that outlook could change if AI investment contracts. For now, the firms arranging shutdowns are continuing to assess and broker data sales.
Further evidence will depend on whether buyers publish results from training with real company records and whether licensing agreements set clearer limits on reuse. The report describes the market as moving toward more formal licensing and vetting, but gives no timetable for common standards. Until more details emerge, the scale of performance benefits, the durability of privacy protections and the distribution of proceeds remain unsettled.
As an affiliate, we earn on qualifying purchases.
Key Questions
What kinds of data are defunct startups selling?
Reported sales include code and workplace archives, such as emails, Slack conversations, documents, calendars and files stored in services such as Google, Notion and Figma. What is included varies by company and transaction.
Who is buying the data?
According to MIT Technology Review, buyers include foundation AI labs, AI-training vendors and established technology companies building models. The report does not identify every buyer or the terms of individual startup sales.
Has the data been shown to improve AI models?
Not publicly, based on the report. It says no company has released comparative tests showing whether real company archives produce better results than synthetic or other training data.
Do workers receive money from these sales?
The report says proceeds typically go to senior creditors and investors, not employees. The distribution can depend on a company’s financial and legal circumstances.
Are the sales legally settled?
No. The report describes open privacy and consent questions, including objections to the Spirit Airlines data sale. Suppliers and buyers say they remove sensitive information, but the extent and effectiveness of safeguards are not uniform or fully clear.
Source: rss
Columbus Day / Indigenous Peoples' Day Picks
long weekend sales
As an affiliate, we earn on qualifying purchases.
