# How to sell your company's data after a shutdown (complete guide)

> Yes, a closed or closing company can often sell or license its old business records (email, team chat, docs, tickets and code) to AI labs. Payouts reported for small startups run from about $10,000 to $100,000, and much more for large archives. The work is in proving you can sign, keeping the data until a deal closes, leaving out what you don't have the right to share, and de-identifying what's left.

Source: https://filefunding.com/learn/sell-company-data-after-shutdown · Updated 2026-10-21

When a company shuts down, its email, team chat, documents, support tickets and code used to be something to delete. In 2026 they are something AI labs pay for. This guide explains who buys, what they pay, what has to be true before you can sell, and the steps from "we're closing" to money in the bank.

> This guide is general information, not legal advice. Data sales touch privacy, employment and contract law. Have your own counsel review any agreement.

## Why AI labs want a closed company's records

The first wave of AI models learned from public text: books, websites, forums. The models labs are building now are meant to do work: answer a support ticket, coordinate a shipment, review code, draft a contract. Public data shows very little of how that work actually happens inside a company. Your archive shows exactly that.

MIT Technology Review reported in October 2026 that buyers now include foundation AI labs, companies that build simulated workplaces for training AI agents, and large tech companies training their own models.[1] One buyer told the magazine that a single company's Slack history, if it captures conversations with enough variety and depth, can be worth more than another company's Slack, email and calendars combined.[1]

The market is young but real:

- **Startups.** Forbes reported that SimpleClosure, a wind-down platform, processed nearly 100 of these deals in a year, paying $10,000 to $100,000 per company.[2] By October, MIT Technology Review put its count above 150 deals, with 47% of its shutdowns now including a code or data sale.[1]
- **A single founder.** One crypto founder told MIT Technology Review he netted roughly $25,000 for his codebase, with about $15,000 more pending for workspace data from Google, Notion, Figma and Slack.[1]
- **A large estate.** In Spirit Airlines' bankruptcy, Google won an August 2026 auction for the airline's de-identified enterprise data with a $10 million bid; Mercor was the backup bidder at $7.5 million, and micro1 later filed notice of a $12.5 million competing bid. Court approval was still pending, with objections from labor groups and contract counterparties on file.[3]

## What you can sell, and what stays out

**What buyers ask for** is records of real work:

- Email threads with customers, vendors and teammates
- Team chat (Slack, Microsoft Teams) where decisions and handoffs happen
- Support tickets and their resolutions
- Docs, wikis, SOPs, checklists and runbooks
- Source code with its history: commits, pull requests and reviews
- CRM activity and sales records
- Call and screen recordings of multi-step tasks

**What stays out** of almost every deal:

- HR files, payroll, medical records and performance reviews. Suppliers routinely exclude HR records.[1]
- Customer payment details and consumer databases
- Material that belongs to someone else: client-confidential documents, data under an NDA, data you licensed from a vendor
- Anything you can't show you have the right to license

You are rarely selling everything. Most deals license a de-identified copy of a defined slice of your records.

## Step 1: Stop the deletion clock

The most common way companies lose this value is simple: the tools get cancelled and the data is deleted before anyone thinks to save it. If you are winding down, keep your main accounts (email, chat, ticketing, code hosting, docs) active, or take full exports before you cancel. Our [startup shutdown data checklist](/learn/startup-shutdown-data-checklist) walks through each system.

How you export matters. Slack's API terms bar apps offered outside your organization from using API data to train a large language model,[4] so chat history should come from the workspace owner's own export, not a third-party tool. Slack's own export options also depend on your plan: on Free and Pro plans, the standard export covers public channels; private channels and direct messages need a Business+ or Enterprise plan.[5]

## Step 2: Confirm who can sign

A buyer will want to know that the person selling has the authority to do it. Who that is depends on how the company is closing:

| Situation | Who usually signs |
|---|---|
| Still operating, or a voluntary wind-down | An authorized officer, often with board approval |
| Dissolved | Whoever is responsible for winding up the company's affairs under state law |
| Chapter 7 or Chapter 11 bankruptcy | The trustee or debtor, with court approval |
| Assignment for the benefit of creditors (ABC) | The assignee |

In a bankruptcy, the Bankruptcy Code adds a privacy rule: if the company's privacy policy restricted transfers of personal information, it can't be sold unless the sale is consistent with that policy or a consumer privacy ombudsman is appointed and the court approves.[6] That is one reason consumer data is usually carved out of these sales, as it was at Spirit, where passenger personal data was set aside for a separate process.[3]

## Step 3: Check what you have the right to license

Before you talk price, sort your records into three piles:

1. **Clearly yours.** Internal email and chat, your own docs and SOPs, your own code.
2. **Strings attached.** Client deliverables, anything covered by an NDA or customer contract, data from vendors, and personal data collected under a privacy policy that didn't contemplate this use. These may need to be excluded or need permission.
3. **Off the table.** HR and medical records, payment data, regulated data you can't de-identify.

At Spirit, two contract counterparties objected to the sale alongside four labor filings.[3] Expect the same questions, at a smaller scale, in any deal. Frankfurt Kurnit's overview of these issues is a good starting point for your lawyer.[7]

## Step 4: Get an estimate, then real offers

Buyers price archives on a few factors: how many people used your systems, how many years of history you have, which tools hold it, how much there is, and your industry. One wind-down firm told MIT Technology Review it values a client's assets by headcount, years in operation, volume of data and how the team used its tools.[1] Our guide to [what your company's data is worth](/learn/what-is-my-company-data-worth) goes through each factor.

Prices are not standardized. Buyers quote "wildly different numbers," according to the buyer MIT Technology Review spoke to.[1] The same article notes that founders often have little visibility into who is buying and little assurance an offer is fair.[1] That is the case for getting more than one offer, and for using someone who knows the buyers.

## Step 5: Scope, de-identify and sign

Once a buyer is interested, the work is mostly paperwork and preparation:

- **Scope.** Agree exactly which systems, date ranges and categories are included, and what is excluded.
- **De-identification.** Names, contact details and other personal identifiers are removed before the data is used for training. At Spirit, each transfer required a de-identification agent to certify the data met California's standard, and the buyer committed not to re-identify it.[3]
- **The agreement.** A license (the buyer gets the right to use a copy) or an outright sale, exclusive or non-exclusive, with permitted uses, restrictions on onward transfer, payment terms and who carries which risks.
- **Delivery.** Usually an export in an agreed format, sometimes processed in a controlled environment rather than handed over raw.

## Where the money goes

If the company is solvent and winding down voluntarily, proceeds go to the company like any other asset sale, and then to creditors and shareholders in the usual order. In a bankruptcy or ABC, they go to the estate. MIT Technology Review notes that proceeds typically flow to senior creditors and investors, not employees.[1] Ask your accountant and lawyer how a sale fits your wind-down before you sign.

## How FileFunding fits

FileFunding doesn't buy your data. We're the transparent middleman: we work with the AI labs and data programs that pay for business records, price and configure your data, and give you a clear comparison of the offers available. Good fits usually have 20+ employees and multiple years of records, such as agencies, managed service providers, e-commerce and other workflow- or CRM-heavy businesses. We take a cut only when a deal closes, disclosed in writing before you sign. You approve every step, and nothing leaves your systems until there's a signed agreement.

Start with your website: we read public information only and show an indicative range in about ten seconds.

## Sources

1. MIT Technology Review, "AI labs are buying data from dead startups", Oct 13, 2026 https://www.technologyreview.com/2026/10/13/1145755/ai-labs-buying-data-from-dead-startups-to-train-models/
2. Gizmodo (summarizing Forbes), "Failed Companies Are Selling Old Slack Chats and Email Archives to Train AI", Apr 17, 2026 https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916
3. Stretto Research Suite, "The Spirit Airlines deidentified data sale: auction results, objections, and the September 30 hearing", Sept 13, 2026 https://chapter11cases.com/blogs/news/the-spirit-airlines-deidentified-data-sale-auction-results-objections-and-the-september-30-hearing
4. Slack, API Terms of Service https://slack.com/terms-of-service/api
5. Slack Help Center, "Export your workspace data" https://slack.com/help/articles/201658943-Export-your-workspace-data
6. 11 U.S.C. § 363 (sale of property; personally identifiable information), Cornell LII https://www.law.cornell.edu/uscode/text/11/363
7. Frankfurt Kurnit Klein & Selz, "Selling company data for AI training: what to know before you say yes" (further reading) https://technologylaw.fkks.com/post/102o455/selling-company-data-for-ai-training-what-to-know-before-you-say-yes