SimpleClosure and Sunset sell dead startups' data to AI labs

MIT Technology Review reports on a market that took shape over the past year: brokers that help failed startups sell their code and internal data to AI companies for use in training.

The opening example is Keelan Miskell, who began winding down his crypto startup this year. He assumed the data was worth little, since the users were anonymous and the company was only a few years old. He has since netted roughly $25,000 for the code base and expects a still-pending amount, likely around $15,000, for the workspace data in Google, Notion, Figma and Slack. The sale was brokered by SimpleClosure. "At least all that stuff we were doing for two and a half years is worth something to someone somewhere," he says.

SimpleClosure belongs to a growing sector serving dying startups. Estimates from Carta and AngelList show shutdowns rising between 26% and 56% annually in recent years, and an end-of-life ecosystem has emerged, including coaches, 'death doulas', and firms that handle legal, tax and operational headaches. SimpleClosure added data sales in April. Its Asset Hub platform helps founders sell code and workspace data such as emails, chats and documents. Buyers include foundation AI labs, vendors that build simulated workspaces for AI training, and mature tech companies building their own models. Asset Hub has handled over 150 deals and expects to return over $10 million to startups by the end of 2026. CEO Dori Yona says such sales, once rare and 'somewhat taboo', are now routine: 47% of SimpleClosure deals include code or data sales.

Rival Sunset also has a rapidly growing data resale business. Founder and CEO Brendan Mahony says Sunset values a dying client's assets by head count, years in operation, volume of data, and how the team used its tools. He compares the material to particle board made from wood shavings nobody thought had value.

Demand emerged suddenly last year, as AI training labs hit the limits of public and synthetic datasets. Surge AI buys data to build reinforcement-learning gyms where clients' AI agents practice tasks such as debugging a complex IT architecture or drafting appellate briefs. Nick Heiner, who leads the company's work on these environments, looks for archives that reflect a range of situations, from disagreements to tricky judgment calls. He says one company's Slack history, if it captures conversations with variety and depth, can be worth more than another's Slack, email and calendars combined.

Real archives avoid common errors of synthetic worlds, such as references to dates that do not exist, but the extent of performance gains remains unclear. No company has yet released tests comparing real-world data training with the alternatives. Pricing reflects that: Heiner says buyers quote wildly different numbers because valuations rest on 'research theses' estimating how many useful tasks come from, say, 100,000 lines of code.

Bigger deals have happened. This past August, in Spirit Airlines' bankruptcy, Google outbid the AI data company Mercor for the carrier's decades of data, including 100 million emails, 500 million Teams messages and 176,000 employee records, at a price of $10 million. Bloomberg reported in September that SpaceX's AI unit is weighing buying data from struggling startups.

The legal picture is murky. Spirit's flight attendants' union objected to the sale on privacy grounds. Attorney Greg Ewing, who advises AI startups, has called such data 'legally radioactive' and points to private details in emails and calendars: doctor appointments, school pickup arrangements, date-night plans. He describes a 'Wild West of AI training' that is only starting to give way to proper licensing agreements and vetted use. Suppliers employ data scientists to scrub personally identifiable information and exclude HR records, and buyers say they scrub too.

There is also the question of who profits. The proceeds typically flow to senior creditors and investors, not employees. Ifeoma Ajunwa, a law professor at Emory University and author of The Quantified Worker, sees this as a form of 'captured capital'. She argues employers should have express consent before selling workers' data, while noting there is no federal law banning such sales.

The article closes with limits for sellers: founders have little to no visibility into who is buying the data and what they plan to do with it, and little assurance of a fair offer. Data prices will likely stabilize in the coming year, it says, unless an AI bubble pops first.

Key facts

• SimpleClosure added data sales in April; its Asset Hub has handled over 150 deals, and 47% of SimpleClosure deals now include code or data sales.
• Founder Keelan Miskell netted roughly $25,000 for his crypto startup's code base, with likely around $15,000 still pending for its workspace data.
• Google paid $10 million for Spirit Airlines' data in the airline's bankruptcy, including 100 million emails, 500 million Teams messages and 176,000 employee records.
• No company has released tests comparing real-world data training with alternatives, so buyers quote wildly different prices.
• Proceeds typically flow to senior creditors and investors rather than employees, and the legal footing is described as shaky.

Why it matters

AI labs have hit the limits of public and synthetic datasets, and demand for defunct companies' internal data emerged suddenly last year. The result is a new revenue line for the startup wind-down business: SimpleClosure says 47% of its deals now include code or data sales, up from a practice once considered 'somewhat taboo'. Shutdowns are also rising between 26% and 56% annually per Carta and AngelList estimates, so the supply of such archives is growing.

Who it affects

Founders winding down companies can recover some value; Miskell's roughly $25,000 for code is one example. Brokers such as SimpleClosure and Sunset gain a new service. Buyers include foundation AI labs, vendors that build simulated workspaces, and mature tech companies training their own models. Employees whose emails, chats and calendars sit in these archives are affected most directly, and they typically do not share in the proceeds.

How to use it

This is a market description rather than a product release. For a founder closing a company, the article points to brokers such as SimpleClosure's Asset Hub and its rival Sunset, which value assets by head count, years in operation, volume of data and how the team used its tools. Heiner says buyers look for archives with variety and depth, such as conversations that include disagreements and tricky judgment calls.

How solid is it

The article is a single reported feature by Matthew King that relies largely on the brokers' own figures, such as SimpleClosure's 150 deals and its expectation of returning over $10 million by the end of 2026, which is a projection. The Spirit Airlines figures are concrete. Miskell's $15,000 is a likely amount, still pending. The article itself says the performance gains from real-world data remain unclear and that no company has released comparison tests.

Risks and caveats

Ewing calls such data 'legally radioactive' and describes a 'Wild West of AI training'. Spirit's flight attendants' union objected on privacy grounds, and Ajunwa says there is no federal law banning such sales of workers' data. Scrubbing of personally identifiable information and HR records is described by suppliers and buyers. Founders have little visibility into who buys the data and little assurance of a fair offer, valuations rest on 'research theses', and prices will likely stabilize in the coming year unless an AI bubble pops first.

“This operational exhaust data is like the wood shavings that everyone thought had no value, but it actually has a ton.”

— Brendan Mahony, founder and CEO of Sunset