Simpleclosure Sells Dead Startups Data to AI Labs

Executive Summary

• SimpleClosure, which began offering data sales in April, has handled more than 150 deals through its Asset Hub platform and expects to return over $10 million to startups by the end of 2026, according to MIT Tech Review AI.
• Keelan Miskell, winding down a crypto startup, netted roughly $25,000 for his code base and a still-pending amount, likely around $15,000, for workspace data in Google, Notion, Figma, and Slack, the report states.
• Google bought Spirit Airlines' data out of bankruptcy this past August for $10 million, outbidding AI data company Mercor for 100 million emails, 500 million Teams messages, and 176,000 employee records, according to MIT Tech Review AI.
• Attorneys and academics quoted in the report raise privacy and consent concerns, and no company has yet released tests comparing real-world data training with alternatives.

Key Takeaways

• A secondary market has formed around liquidating the digital remains of failed startups, with brokers, buyers, and pricing methods emerging in roughly a year.
• Demand intensified as AI training labs reached the limits of public and synthetic datasets, pushing buyers toward real operational archives.
• Pricing remains unsettled because no one has published comparative performance tests, so valuations rest on buyer research theses rather than measured results.
• Legal exposure is unresolved. A union objection, attorney criticism, and the absence of a federal ban define the current environment.

MIT Tech Review AI Examines How Shutdown Data Became an AI Input

Startups fail constantly. What changed recently is that their residue—code, emails, chats, documents, calendars—has acquired a buyer base willing to pay real money for it.

MIT Tech Review AI traces the shift to SimpleClosure, an end-of-life services firm that added data sales in April through a platform called Asset Hub. Founders list code and workspace data; buyers include foundation AI labs, vendors that build simulated workspaces for training, and mature technology companies developing their own models.

SimpleClosure chief executive Dori Yona told the publication that these transactions, once rare and somewhat taboo, have become routine. Forty-seven percent of the company's deals now include code or data sales. Competitor Sunset runs a data resale business of its own, and founder and CEO Brendan Mahony says the firm values a dying client's assets by head count, years in operation, data volume, and how the team used its tools.

The analogy Mahony offers is manufacturing scrap. Operational exhaust data, he says, is like wood shavings everyone assumed were worthless until someone found a use.

MIT Tech Review AI Reports Why Buyers Want Real Archives

The demand appeared suddenly last year, per the report, when AI training labs hit the ceiling of public and synthetic datasets. Surge AI buys data to construct what it calls reinforcement-learning gyms, where clients' AI agents practice tasks such as debugging a complex IT architecture or drafting appellate briefs.

Nick Heiner, who leads that work at Surge AI, says he looks for archives reflecting a range of situations, from disagreements to tricky judgment calls. Variety and depth matter more than raw volume. Heiner's example: one company's Slack history, if rich enough, can be worth more than another firm's Slack, email, and calendars combined.

The stated advantage of real archives is error avoidance. Synthetic worlds can reference dates that do not exist. MIT Tech Review AI is careful to note the limit here: the extent of performance gains from real-world data training remains unclear, and no company has released tests comparing it with alternatives.

Related: Launching an AI in Education Startup in 2026: Top 5 AI Automation Tips

Pricing reflects that uncertainty. Buyers quote widely different numbers, Heiner says, because valuations rest on research theses estimating how many useful tasks emerge from, say, 100,000 lines of code.

MIT Tech Review AI on the Spirit Airlines Deal and Its Scale

The largest documented transaction in the report came out of a bankruptcy. This past August, as part of Spirit Airlines' restructuring, Google outbid the AI data company Mercor to acquire the carrier's decades of data: 100 million emails, 500 million Teams messages, and 176,000 employee records. The price was $10 million.

The report also notes that, as reported elsewhere, SpaceX's AI unit is weighing buying data from struggling startups. MIT Tech Review AI treats this as a signal of buyer interest rather than a completed transaction.

Individual deals run far smaller. Miskell's roughly $25,000 for code and a pending amount likely near $15,000 for workspace data illustrates the long tail. SimpleClosure's aggregate expectation, over $10 million returned by the end of 2026, shows how those small deals scale when volume is high.

For deeper context, see our Fintech analysis: "ElevenLabs and Universal Music Group Announce AI Voice Licensing Agreement".

MIT Tech Review AI Identifies the Privacy and Consent Fault Lines

The legal ground is unsettled. Spirit's flight attendants' union objected to the sale on privacy grounds. Greg Ewing, an attorney who advises AI startups, has called such data legally radioactive.

Suppliers employ data scientists to scrub personally identifiable information and exclude HR records, and buyers say they scrub as well. Ewing counters that private details persist in emails and calendars—doctor appointments, school pickup arrangements, date-night plans. He describes a Wild West of AI training only now giving way to licensing agreements and vetted use.

Distribution of proceeds is a second fault line. MIT Tech Review AI reports that windfall proceeds typically flow to senior creditors and investors, not employees. Ifeoma Ajunwa, a law professor at Emory University and author of The Quantified Worker, frames this as captured capital. Monitoring to help workers do their jobs has legal precedent, she says, but that justification disappears once a firm is out of business. Ajunwa argues employers should obtain express consent before selling workers' data, while acknowledging no federal law bans such sales.

MIT Tech Review AI on Limits Facing Founders and the Market

The report is explicit that this market has constraints. Founders have little to no visibility into who buys their data or what buyers plan to do with it, and little assurance the offer is fair.

Additional coverage: Merck and Moderna mRNA Cancer Vaccine Beats KEYTRUDA Alone in Phase 3 Melanoma Trial

MIT Tech Review AI expects data prices to stabilize in the coming year unless an AI bubble pops first. Either outcome leaves the appraisal function intact, with the emerging class of tech undertakers positioned to value whatever remains.

MIT Tech Review AI Signals Table

Entity	Recent Focus	Geography	Source
SimpleClosure	Asset Hub platform selling code and workspace data; added data sales in April; over 150 deals; expects over $10 million returned by end of 2026	Not specified in source	MIT Tech Review AI
Sunset	Rapidly growing data resale business; values assets by head count, years in operation, data volume, tool usage	Not specified in source	MIT Tech Review AI
Surge AI	Buys data to build reinforcement-learning gyms for AI agent task practice	Not specified in source	MIT Tech Review AI
Google	Bought Spirit Airlines data out of bankruptcy in August for $10 million, outbidding Mercor	Not specified in source	MIT Tech Review AI
Spirit Airlines	Bankruptcy estate sold decades of data including 100 million emails, 500 million Teams messages, 176,000 employee records	Not specified in source	MIT Tech Review AI

The source does not provide geographic detail for any entity listed, so that column is marked accordingly rather than inferred.

What This Means for Practitioners

For founders, buyers, and procurement teams, the practical shift is that data assets now have a secondary market with brokers, valuation heuristics, and a growing record of completed deals. That does not make pricing reliable. Without published comparative tests of real-world versus synthetic training data, any valuation rests on a buyer's thesis about task yield, and quoted prices vary widely for that reason. Teams contemplating a sale should assume limited visibility into buyer identity and downstream use, and should treat consent, scrubbing, and record-exclusion practices as negotiable terms rather than settled defaults.

MIT Tech Review AI Implementation Risks

The risks documented in the report are legal and reputational rather than technical. Data described as legally radioactive can carry privacy claims, as the flight attendants' union objection to the Spirit sale indicates. Scrubbing personally identifiable information and excluding HR records are stated supplier and buyer practices, not verified guarantees, and attorney commentary points to residual private detail in emails and calendars. Proceeds typically bypass employees, and Ajunwa notes no federal law bars these sales, leaving policy to develop unevenly. Pricing risk compounds legal risk: with no comparative performance tests published, buyers are paying against estimates, and the report expects prices to stabilize within the coming year unless an AI bubble pops first.

Editorial independence disclosure: this analysis is based solely on reporting published by MIT Tech Review AI and does not reflect any relationship with the companies named.

Source note: all facts in this article derive from MIT Tech Review AI's report on AI labs buying data from dead startups, published October 13, 2026.