Your company’s old Slack messages might be worth more than the laptops everyone wrote them on. Right now, data brokers are paying anywhere from $100,000 to $2 million for workplace archives, and the pitches landing in inboxes come with 48-hour deadlines.
WIRED’s Steven Levy just published the receipts. Someone offered media companies between $350,000 and $1.8 million for their full work records, documents, emails, Slack logs, support tickets, the lot. A second pitch added call transcripts and bumped the ceiling to $2 million. If you’ve ever wondered whether sell data to AI companies is a real market or spam, it’s real, it’s huge, and it’s moving faster than the rules around it.
Why AI labs pay millions for your old emails
The simple version: public internet data ran out. Labs have scraped everything worth scraping, and synthetic data only takes you so far.
What labs actually want now is a record of how real work happens. Not the polished final report, but the messy middle. The argument with a client, the three revisions of a proposal, the Slack thread where a team diagnosed a bug at 2am. MIT Technology Review spoke with Nick Heiner at Surge AI, a company building “reinforcement-learning gyms” where AI agents practice on archived reality. His take: one company’s Slack history, if it captures real variety and depth, can be worth more than another company’s Slack, email, and calendars combined.
Anthropic and OpenAI sit at the end of this pipeline, and according to the pitches quoted by WIRED, the labs’ total spend on private data has passed $45 billion. Brokers are the middlemen packaging companies’ archives into training fuel.
The price list, from real deals
If you’re weighing whether to sell data to AI companies, here’s the table worth bookmarking:
| Deal type | What sold | Price |
|---|---|---|
| Micro1 data program | Midsize company’s full archive (Slack, CRM, email) | $100,000-$2,000,000 |
| WIRED-reported pitch | Media company work records, one-time fee | $350,000-$1,800,000 |
| Google (Spirit Airlines bankruptcy) | 100M emails, 500M Teams messages, 176K records | $10,000,000 |
| SimpleClosure dead-startup sale | Code base from winding-down startup | ~$25,000 |
| Dead startup workspace | Google, Notion, Figma, Slack archives | ~$15,000 |
Two patterns jump out. Bigger and more complete beats bigger alone. Brokers like Micro1 generally want the whole corpus, because the value is how the pieces connect. And specialized work beats generic work. A law firm’s or accounting shop’s records sell at a premium since AI developers are racing to automate exactly those fields.
Micro1 disclosed plans to spend $1 billion on data licensing over the next year. SimpleClosure, which helps startups wind down, has handled over 150 data deals and expects to return more than $10 million to founders by the end of 2026. Nearly half its deals now include code or data sales, up from a time when selling archives was considered outright taboo.
How selling actually works
The mechanics to sell data to AI companies are simpler than you’d think, and the methods matter if you’re evaluating an offer.
Brokers like Micro1 operate three ways. First, raw licensing: they take your data, scrub it to protect privacy, and license it to labs. Second, agent feedback: they hand companies an in-progress AI agent and route your feedback to developers. Third, simulation: your archive gets frozen at a point in time and becomes an “environment” where models learn to work like your employees did.
The VanHack founder Ilya Brotzky told Business Insider he didn’t believe the first offer was real. He eventually sold his Google suite, Confluence, Slack, and CRM data after digging into the privacy process. That’s the right order of operations, by the way: investigate the anonymization pipeline before you sign anything, not after.
What determines your price? Volume matters, since data is often priced by the terabyte. Completeness matters more. Exclusivity and how unusual your industry’s work is can swing the number the most. And there’s a market for leftovers too: founders winding down companies are converting dead archives into five-figure payouts instead of deleting them.
The 48-hour offer and other red flags
Now the part the pitch emails skip.
Business Insider reports owners are getting “exploding offers,” six-figure sums that expire in 48 hours, plus Instagram ads claiming “AI needs your data” and promising millions. Pressure is a sales tactic, not a market signal. Any pitch to sell data to AI companies that rushes you past legal review is telling you something.
Privacy risk is the deeper problem. Employees usually don’t know their workplace communications are for sale. Anonymization isn’t foolproof: Meta suspended an employee-data collection program after supposedly anonymized records, private conversations included, leaked internally. During the Spirit Airlines sale, flight attendants formally objected on privacy grounds, and attorney Greg Ewing called this material “legally radioactive,” a Wild West that proper licensing is only starting to tame.
Think about what lives in work email: doctor appointments, school pickups, salary gripes, health scares. Scrubbing catches most of it. Most.
There’s a strategic cost too, and it might be the biggest one. A company’s Slack archive is its playbook, how its people think, decide, and solve problems. Selling it for a one-time check is selling your seed corn. The buyer gains a simulation of your edge, and you gain a number that depreciates the moment it clears.
If you still want to explore it
Some situations make sense. Winding down a company with archives you’d otherwise delete? Selling beats shredding, and SimpleClosure exists precisely for that. Own a data-heavy business with genuinely unique workflows, healthy cash flow, and no intention of ever competing on process secrecy? A seven-figure fee can fund the next stage. Just route everything through counsel, demand the anonymization spec in writing, and keep HR records and anything regulated out of the deal entirely.
And if your company isn’t for sale but you personally want control over what AI sees from your life, that battle is smaller and winnable. Our walkthrough of the ChatGPT training setting shows how to check whether your chats train models. Why labs are also hunting fresh data without the public internet is its own story, which we covered in AI training without internet. Even websites are fighting back, with major platforms blocking AI agents from crawling their content.
The takeaway
The market to sell data to AI companies is real, liquid, and priced like it: five figures for a dead startup’s leftovers, seven figures for a living company’s playbook, ten for an airline’s whole memory.
If you own something valuable, slow down and lawyer up before signing anything, because the 48-hour clock is their tool, not yours. If you’re just an employee, ask the question that makes compliance people move fast: has our data been offered to a broker? The answer will tell you more about your company than any all-hands meeting.
Takeaway: Brokers pay $100K to $2M for workplace archives, but the exploding offer in your inbox deserves a lawyer, not a signature. Founders winding down should check SimpleClosure’s model first, everyone else should ask legal one direct question this week: has anyone pitched us on our Slack and email data? Your archive is either an asset you sell carefully or an edge you keep. Just decide before an Instagram ad decides for you.