
Monetize
your company's data.
We connect untapped data with AI labs.
We run the deal from first call to payment
You bring the data. Everything between you and the buyer is our job.
Value
A short call about what you hold. You get a free value range, with nothing uploaded.
Clear
We check what you have the rights to license and what has to come out first.
Prepare
Personal data is removed and the rest is cleaned to the standard buyers expect.
Match
An anonymous description goes to every buyer it fits, so they compete on price.
Close
You pick the offer, sign a license you've read, and get paid.
Why did Google pay $10 million for a bankrupt airline's old emails, files and CRM?
The bidding opened at $5M. It closed at twice that. To see why, you have to look at how AI gets built.
Every new AI model is trained on more data than the last.
Labs scraped everything public to get it: the web, books, code, forums.
300B
tokens · GPT-3, 2020
15T
tokens · Llama 3, 2024
Now the public internet is running out.
Scraping more of the same web won't build a better model.
So the labs are paying for data nobody else has.
Emails, code and records from game studios, mom-and-pop shops, robotics companies. Anyone.
- News Corp→OpenAI$250M+
- Reddit→Google$60M
- Shutterstock→Big tech$25–50M
Reported licenses. Sources
But the bottleneck is bringing private data to market.
Most companies don't know their data is worth anything, or who would buy it.
So we became the broker.
We find the companies sitting on this data, clear it for sale and take it to the labs that will pay for it.
What we do
1/Your company's data is worth more than you think
The work your team has already done is what the labs are short of. We find the parts of it that sell, and the buyers who pay the most for them.
Codebases
Repositories with their history: commits, reviews and issues that show why the code is the way it is.
- GitHub
- GitLab
- Bitbucket
- Perforce
Messages
The threads where your team works things out, including the decisions and the reasoning behind them.
- Slack
- Teams
- Gmail
- Outlook
Documents
Docs, reports, specs and decks: the written record of what your company knows.
- Google Drive
- SharePoint
- Notion
Business records
Exports from the systems that run the business, like pipelines, invoices and orders.
- Salesforce
- HubSpot
- QuickBooks
Tickets and tasks
Work captured end to end: the request, the back-and-forth and the finished result.
- Jira
- Zendesk
- Asana
Audio, video and images
Recordings, footage and image libraries your company owns the rights to.
- Calls
- Playtests
- Asset libraries
2/Game and software studios: license your source code
Labs training coding models want real codebases, written by real teams over years. A shipped game or product can be licensed without giving up ownership of it.
What buyers look for
Original code
Written by your team, not generated or copied in.
Real history
Commits, reviews and issues that show how it was built.
Clear ownership
Someone who can sign for the rights to license it.
The license covers a copy for training. Your code stays yours, and nothing is shared without an NDA and your sign-off.
- ProjectAurora/UE 5
- Source/C++ · 6 yrs history
- Content/Blueprints/graphs
- Content/Maps/levels
- Design/GDD + tickets
Sample project.
3/For AI labs: private data you can't scrape
We bring licensed data from companies that don't sell it anywhere else, and handle the sourcing, rights and paperwork on their side.
Sourced to your spec
Tell us the gap in your model. We find the companies whose data fills it, including ones that have never sold before.
Rights cleared first
Every dataset comes with a seller who can sign for it and a license that says what you can do with it.
Delivered clean
Personal data removed, formats normalized, and samples under NDA before you commit.
4/Rights and privacy come before price
Selling data comes with legal requirements on both sides. We check the rights, strip personal information and run the NDA and license in order, so none of it is on you to set up.
Personal data stays out
Names, contact details and anything identifying are removed before data reaches a buyer.
Anonymous until you say so
Buyers see a description of your data, not your name, until you choose to go further.
NDA before any sample
No buyer sees a redacted preview without a signed NDA.
Licensed, not handed over
Most deals license a copy for a set purpose. You keep the original and the rights you don't sell.
You approve every step
Which buyers, which offer, which terms. Nothing is shared or signed without you.
Paid when you are
Nothing up front. Our fee comes out of a closed deal, and you see it in writing before you sign.
What data sells for
Median public prices by type of data, drawn from 1,487 public prices. A valuation starts here and moves with size, age and rights.
Median price by data type
- Company records$276,000 per company
- Source code$2,000 per repository
- Text and archives$332 per title
- Datasets and tasks$300 per task
- Audio and speech$135 per hour
- Video$105 per hour
- Expert work$80 per hour
- Images$0.57 per image
What buyers say they pay per company
- Mode (BizBuyData)$100K to $5M
- Handshake AI$100K to $4M
- micro1$100K to $2M
- Polyshares$100K to $2M
- Mercor$10K to $1.5M
- Scale AI$10K to $1M
Every bar uses the same $0 to $5M scale.
Free guides and tools
Start with what people search for most: how data deals work, the terms behind them and free tools for each step.
Data types
Data broker laws
- California data broker registry and the Delete Act
- Vermont data broker registry
- Texas data broker registry
- New Jersey data broker law
Questions about selling your data
Short answers to what companies ask before a valuation.
What does DataBroker do?
DataBroker is a data broker. We help companies license the data they already hold, such as codebases, documents, messages and business records, to AI labs and other buyers. We value it, check the rights, remove personal information, find the buyers and run the deal through to payment.
Is it legal to sell my company's data?
Yes, when your company holds the rights to it and the deal follows privacy law. Most deals are licenses: the buyer gets to use a copy for an agreed purpose, and you keep the original. Before we take anything to a buyer, we check what you can license and what has to come out first. This is general information, not legal advice.
Will personal data about my customers or employees be sold?
No. Personal information is removed or de-identified before anything reaches a buyer, and buyers only see a description and a redacted preview under NDA until a license is signed. Data that can't be cleaned to that standard stays out of the deal.
Do I have to upload anything to get a valuation?
No. A valuation starts with a short call about what kinds of data you have, roughly how much and over how many years. Nothing leaves your systems until a buyer is seriously interested and you've agreed to go ahead.
What does it cost?
The valuation is free and there's nothing to pay up front. We're paid when a deal closes, and you see our fee in writing before you sign anything.
Would labs buy our game's or product's source code?
Often, yes. Labs training coding models value original code with a real development history behind it. Tell us roughly what the codebase is, how big it is and how long it's been in development, and we'll tell you where it stands.
I work at an AI lab. Can you find data for us?
Yes. Tell us what your models are missing and the rights you need, and we'll source it from companies that hold it, with the licensing and preparation handled on their side.
Find out what your data is worth
A free valuation, and no commitment to sell.