# Matthew King argues in MIT Technology Review that failed startups’ data has a new market

[Canonical reading](https://escapemeridian.com/readings/matthew-king-argues-in-mit-technology-review-that-failed-ed6c746e)

Matthew King’s article was published on October 13, 2026. He describes a growing market in which firms selling services to dying startups broker sales of their code and workspace data to AI labs and other buyers. Real company archives can contain varied conversations and work situations for training AI systems. Nick Heiner of Surge AI says one company’s Slack history can be more valuable than another’s larger collection if it captures greater variety and depth. But no company has released tests comparing performance gains from training on real-world data with alternatives. Buyers’ prices reflect uncertain estimates of the useful tasks those archives might yield.

The sales can return money to failed businesses, but proceeds typically go to senior creditors and investors, not employees. They also raise privacy and consent concerns: Spirit Airlines’ flight attendants’ union objected to the sale of employee data, and law professor Ifeoma Ajunwa argues employers should get workers’ express consent. Founders may have little visibility into buyers’ identities and plans, while there is little assurance they receive fair offers. King closes by predicting that data prices will likely stabilize in the coming year, unless an AI bubble pops first; he says tech’s new undertakers will be there to appraise what remains.

Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.

## Original source

MIT Technology Review · Article

Original source date: Oct 13, 2026 (2026-10-13T09:00:00Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-20T04:33:09.953Z. First observed: 2026-11-20T02:32:05.838Z. Retrieved: 2026-11-20T02:34:00.715Z.

- [AI labs are buying data from dead startups to train their models](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Limits

Text as retrieved on November 20, 2026; originally published October 13, 2026. The article says no company has released comparative tests, so the extent of performance gains from training on real-world data remains unclear. The article does not give a basis for assessing whether data sellers receive fair prices. The article says founders have little assurance they do.


## Open question

How often do buyers’ claimed training benefits justify the prices paid, given that no company has released comparative tests?


## Source-stated

Matthew King reports that SimpleClosure brokers sales of startup code and workspace data to AI labs and other buyers.

> Buyers include foundation AI labs, vendors that construct simulated workspaces for AI training, and mature tech companies building their own models.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

Nick Heiner says the value of company archives depends on the variety and depth of their work conversations.

> 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.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King reports that comparative tests of training with real-world data and alternatives have not been released.

> No company has yet released tests comparing results of real-world data training with the alternatives.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King reports that sales proceeds typically go to senior creditors and investors, not employees.

> The proceeds typically flow to senior creditors and investors—not employees.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

Ifeoma Ajunwa says employers should get workers’ express consent before selling their data.

> “Employers should have express consent before selling workers’ data,” says Ajunwa

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King says founders have little visibility into who buys the data and what buyers intend to do with it, and little assurance of a fair offer.

> Founders have little to no visibility into who is buying the data and what they plan to do with it. Plus, there’s little assurance they’re getting a fair offer.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King predicts data prices will likely stabilize in the coming year, unless an AI bubble pops first.

> Data prices will likely stabilize in the coming year, though, unless an AI bubble pops first.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source and contributor context

MIT Technology Review · Publisher

The publisher is identified here. No individual author has been inferred.


## Source media

Media retrieval status: unchecked.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Editorial provenance

Version 1. Publication ID: 027efe21-06f8-d27f-d238-68b2f9c8f200.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-20T04:29:15.344Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical.

Observation: f535d7a4c37f0d2379a68981f535599089567e5f16da0191fcaf7c5e1a1992ac. Source text hash: a096905fdc6a1f32219d39b8f0457c7e82b990d38ccbf542889239802bb8a9ce.


## Revision and source-check history

### Version 1 — 2026-11-20T04:33:09.953Z

Matthew King’s article was published on October 13, 2026. He describes a growing market in which firms selling services to dying startups broker sales of their code and workspace data to AI labs and other buyers. Real company archives can contain varied conversations and work situations for training AI systems. Nick Heiner of Surge AI says one company’s Slack history can be more valuable than another’s larger collection if it captures greater variety and depth. But no company has released tests comparing performance gains from training on real-world data with alternatives. Buyers’ prices reflect uncertain estimates of the useful tasks those archives might yield.

The sales can return money to failed businesses, but proceeds typically go to senior creditors and investors, not employees. They also raise privacy and consent concerns: Spirit Airlines’ flight attendants’ union objected to the sale of employee data, and law professor Ifeoma Ajunwa argues employers should get workers’ express consent. Founders may have little visibility into buyers’ identities and plans, while there is little assurance they receive fair offers. King closes by predicting that data prices will likely stabilize in the coming year, unless an AI bubble pops first; he says tech’s new undertakers will be there to appraise what remains.

## Original source

MIT Technology Review · Article

Original source date: Oct 13, 2026 (2026-10-13T09:00:00Z).

Date basis: Date from the publisher’s own records, through its API.

Added to Escape Meridian: 2026-11-20T04:33:09.953Z. First observed: 2026-11-20T02:32:05.838Z. Retrieved: 2026-11-20T02:34:00.715Z.

- [AI labs are buying data from dead startups to train their models](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Limits

Text as retrieved on November 20, 2026; originally published October 13, 2026. The article says no company has released comparative tests, so the extent of performance gains from training on real-world data remains unclear. The article does not give a basis for assessing whether data sellers receive fair prices. The article says founders have little assurance they do.


## Open question

How often do buyers’ claimed training benefits justify the prices paid, given that no company has released comparative tests?


## Source-stated

Matthew King reports that SimpleClosure brokers sales of startup code and workspace data to AI labs and other buyers.

> Buyers include foundation AI labs, vendors that construct simulated workspaces for AI training, and mature tech companies building their own models.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

Nick Heiner says the value of company archives depends on the variety and depth of their work conversations.

> 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.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King reports that comparative tests of training with real-world data and alternatives have not been released.

> No company has yet released tests comparing results of real-world data training with the alternatives.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King reports that sales proceeds typically go to senior creditors and investors, not employees.

> The proceeds typically flow to senior creditors and investors—not employees.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

Ifeoma Ajunwa says employers should get workers’ express consent before selling their data.

> “Employers should have express consent before selling workers’ data,” says Ajunwa

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King says founders have little visibility into who buys the data and what buyers intend to do with it, and little assurance of a fair offer.

> Founders have little to no visibility into who is buying the data and what they plan to do with it. Plus, there’s little assurance they’re getting a fair offer.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source-stated

King predicts data prices will likely stabilize in the coming year, unless an AI bubble pops first.

> Data prices will likely stabilize in the coming year, though, unless an AI bubble pops first.

- [Supporting source](https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/)

## Source and contributor context

MIT Technology Review · Publisher

The publisher is identified here. No individual author has been inferred.


## Source media

Media retrieval status: unchecked.


## Terms and references

Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.


## Editorial provenance

Version 1. Publication ID: 027efe21-06f8-d27f-d238-68b2f9c8f200.

Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.

Source check: passed by anthropic/claude-opus-5.5 on 2026-11-20T04:29:15.344Z. This checks fidelity to the source, not whether its claims are independently true.

Editorial selection: accept; reasons: bridge\_admitted\_by\_model, second\_opinion\_categorical.

Observation: f535d7a4c37f0d2379a68981f535599089567e5f16da0191fcaf7c5e1a1992ac. Source text hash: a096905fdc6a1f32219d39b8f0457c7e82b990d38ccbf542889239802bb8a9ce.


### passed — 2026-11-20T04:29:15.344Z — anthropic/claude-opus-5.5

Publication: 027efe21-06f8-d27f-d238-68b2f9c8f200

attribution — pass: The reading correctly attributes the article to Matthew King at MIT Technology Review. Quotes and claims are attributed to Heiner, Ajunwa and the union as in the source. 'Argues' in the title is slightly strong for a reported feature, but it does not misattribute anything.

> Matthew King is a technology and environmental journalist based in Baltimore.

summary — pass: The summary clearly explains the market, the uncertainty about performance gains and pricing, who gets the proceeds, the privacy and consent concerns, and King's hedged closing forecast. The Spirit union reference lacks the context of Google's purchase of the airline's data, but this does not block understanding. The plain-language framing is accessible without specialist background.

> Now tech undertakers are selling the data remains to AI labs.

date — pass: The stated date of October 13, 2026 matches originalPublished. The modifiedAt timestamp predates publication, so it does not indicate a later edit. The reading does not use the related-story sidebar. The forecast is reported as King's prediction, and no archive disclosure is needed for a publisher page.

> Data prices will likely stabilize in the coming year, though, unless an AI bubble pops first.

support — pass: Each claim matches its verbatim excerpt. 'Can return money to failed businesses' is supported by Miskell's sale and SimpleClosure's figures. The question follows from the stated absence of comparative tests and does not assume an answer.

> No company has yet released tests comparing results of real-world data training with the alternatives.

media — pass: No media was selected and no verified candidates were supplied, so nothing is misidentified.

## Recorded references

Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter.

No public references recorded here yet.

[Reference data (JSON)](/readings/ed6c746eb2ac18681f801ab4892a7adb7038b6c7e066fc917abf1f3db0c31511/references.json)

## Corrections and challenges

[Send a correction or an opposing source](https://escapemeridian.com/corrections/)

[JSON record](https://escapemeridian.com/readings/matthew-king-argues-in-mit-technology-review-that-failed-ed6c746e.json)
