Selling data
Selling a company's workspace data to AI labs
Sell workspace data only after you know who buys it, what deals have paid, and what has to be stripped out before anyone sees it.
Published
The short answer
AI labs and their suppliers now buy companies' workspace data, meaning chats, emails, documents, calendars and code, mostly from startups that are closing. Published deals range from about $25,000 for one startup's codebase to $10 million for Spirit Airlines' workplace archive. The data must be yours to sell, and personal and HR information has to come out first.
What AI labs buy from operating companies
Labs training models to do office work need examples of real work: how a team argues in a chat thread, drafts and revises a document, files a bug and fixes it. Public and synthetic data run short on that, which MIT Technology Review says is why demand for defunct companies' data appeared suddenly in 2025.
The buyers it names are foundation labs, vendors that build simulated workspaces for training agents, and large tech companies building their own models. Surge AI, for one, buys data for reinforcement-learning gyms where agents practise tasks.
What they buy
- Code bases, with their history.
- Chats from tools such as Slack and Microsoft Teams.
- Email and calendars.
- Documents and design files from tools such as Google Workspace, Notion and Figma.
- Employee records, in at least one bankruptcy sale, which is exactly what most suppliers now say they exclude.
Published workspace data deals and prices
- Google paid for Spirit Airlines' workspace data in bankruptcy
- $10 million
- for one shut-down crypto startup's codebase
- About $25,000
- closing companies' assets sold through SimpleClosure
- 150+
- of SimpleClosure's deals now include code or data
- 47%
| Seller | What was sold | Buyer | Price | Source |
|---|---|---|---|---|
| Spirit Airlines (in bankruptcy) | About 100 million emails, 500 million Teams messages and 176,000 employee records | Google, outbidding Mercor | $10 million | MIT Technology Review |
| A shut-down crypto startup (founder Keelan Miskell) | Its codebase | Not named | About $25,000, plus a pending payment of roughly $15,000 for Google, Notion, Figma and Slack data | MIT Technology Review |
| Failed startups | Production code bases, cleaned and then licensed or resold | Turing | Reportedly tens of thousands of dollars each | Let's Data Science, December 2025 |
| Over 150 closing companies | Assets sold through SimpleClosure's Asset Hub; 47% of its deals now include code or data | Various | SimpleClosure expects to return over $10 million to startups by the end of 2026 | MIT Technology Review, quoting CEO Dori Yona |
Buyers quote very different prices for similar data. Surge AI's Nick Heiner told MIT Technology Review that valuations rest on research theses about how many useful tasks a given volume holds, such as per 100,000 lines of code.
How buyers value it
The resale firm Sunset values assets by headcount, years in operation, data volume and how the team used its tools. For code, our codebase value calculator turns those kinds of factors into a rough range.
Before you sell workspace data: rights and privacy
- 1
Confirm the company owns it
Workspace data sits in third-party tools under their terms, and much of it was written by employees, contractors and customers. Check your contracts, the tools' terms and your contributor agreements before promising anything.
- 2
Decide what is out of scope
HR files, health information, customer personal data and anything under a confidentiality agreement with someone else. Suppliers told MIT Technology Review they exclude HR records and scrub identifying information.
- 3
Scrub, then check the scrub
Names, emails, phone numbers, keys and passwords hide in chats and calendars. Run a scan for personal data and secrets, then sample by hand: one lawyer quoted in the reporting called this data legally radioactive.
- 4
Sign before sharing
Put a non-disclosure agreement in place before a buyer sees samples, and agree in writing what they may use the data for and whether they may resell it.
- 5
Tell the people affected where you can
There is no federal law banning the sale of workers' data, but a law professor quoted by MIT Technology Review argues employers should get express consent, and Spirit's flight attendants' union objected to its sale. Expect the question.
Sources
2 sources · checked Nov 9, 2026
The DataBroker team checked the facts on this page against 2 public sources, last on .
Selling a company's workspace data to AI labs FAQ
Can a startup sell its Slack data to AI companies?
Some have, usually at shutdown, but only after checking that the company may transfer it and removing personal and HR information. Ownership of messages written by employees and customers is the first thing to settle.
How much is a startup's workspace data worth?
Published examples run from about $25,000 for a small startup's codebase to $10 million for Spirit Airlines' archive. Buyers value data by how many useful training tasks it holds, so prices vary widely.
Who buys company data for AI training?
Foundation labs, vendors that build simulated workspaces and training environments, and large tech companies training their own models, often through intermediaries such as SimpleClosure and Sunset.
Is it legal to sell employee emails and chats?
There is no federal law banning it, but privacy, employment and contract rules can apply, and unions have objected. Remove personal and HR information and take legal advice before selling.
Related
- How to sell data to AI companies
Sell data to AI companies only after you can prove where it came from, that you may license it, and why it isn't online already.
- How to sell software you have built
How to sell software you have built, whether that means the whole product, the IP, or licensing the codebase as training data.
Codebase value calculatorA codebase value calculator that reads a public GitHub repository in your browser and estimates rebuild effort and cost with basic COCOMO.
PII scannerA PII scanner that runs in your browser: paste text or drop a file to see personal data and leaked secrets before you share it.
Free tool
Try the PII scanner
A PII scanner searches text for personally identifiable information, such as email addresses, phone numbers and government ID numbers, so it can be removed or protected before the text is shared.
- Free
- No sign-up
- Runs in your browser