An AI Company Wants to Buy Your Data. What Should You Do?

Someone has contacted your business offering money for your company data. Maybe they want your emails, Slack messages, customer records, or internal documents.

Is the offer legitimate? How much could your data be worth? And what would you be giving up?

If you've just received an offer, do these three things first:

1. Get the details in writing. Ask who is buying, what records they want, and what they're offering.

   Once you have written terms, use the free AI data licensing offer comparison Sheet to compare the money, rights, and unanswered questions.

2. Don't share any data yet. You can explore the opportunity without sending files or connecting your systems.

3. Find out who will use the data. The company contacting you may be collecting it for someone else. Ask who ultimately receives it and what they're allowed to do with it.

Here's what you should know before moving forward.

Are AI companies really buying business data?

Yes. Several companies now have dedicated programs to pay businesses for access to their internal records.

They're interested in things like customer conversations, project histories, support tickets, and internal documents. These records show how people solve problems, make decisions, and get work done.

That can be valuable for training and testing AI systems.

Three companies actively advertising these opportunities are:

Scale AI looks for operational data from tools like Slack, Salesforce, Jira, and Google Workspace. It says it generally prefers businesses with at least 40 employees, although it considers smaller companies individually.

Mercor licenses business records, prepares them for AI training, and supplies the resulting datasets to AI labs. Its program covers industries including manufacturing, professional services, finance, and logistics.

micro1 looks for internal documentation, CRM records, project histories, and examples of business decisions. Its enterprise data partnership program targets companies with at least 30 employees.

Comparing two of these programs? See micro1 vs Mercor, micro1 vs Scale AI, Mercor vs Handshake AI, Mercor vs Troveo, and micro1 vs Handshake AI.

These aren't just opportunities for huge corporations. Established small and mid-sized businesses may have exactly the kinds of records these companies want.

But having valuable data doesn't automatically mean you'll receive an offer. Buyers assess the quality, quantity, uniqueness, and usefulness of what you have.

1. Who is actually making the offer?

The company that contacts you might not be the company that ultimately uses your data.

Some companies collect and prepare data for AI labs. Others introduce business owners to potential buyers.

There's money in making those introductions, too.

Scale AI and micro1 both advertise referral payments of up to $50,000 for successful introductions.

So if someone unexpectedly contacts you about your company's data, they could be an employee, a sales representative, or an independent referral partner.

Ask who they represent, who would sign the agreement, and who would pay you. If they're arranging an introduction, ask how they're compensated.

Verify the company and your contact using an independently found website or contact method.

If you're dealing with a newer or unfamiliar company, check its business registration, leadership, previous work, and ability to meet its financial obligations. Ask for references and details about its security practices.

Before signing anything, you should know who you're dealing with and who will ultimately receive your data.

2. What data do they want?

An offer for "your company data" is too broad to evaluate.

Ask which systems, records, and years of history the buyer wants. Do they want a one-time copy of your existing records or ongoing access to new information?

For example, your company might have eight years of customer support tickets, including the original customer requests, internal discussions, and final resolutions.

You can describe that history and its approximate size without sharing the records themselves.

You also need to check what you're allowed to share.

Your systems may contain confidential customer information, employee messages, supplier documents, or material covered by third-party agreements.

Review the relevant contracts and privacy obligations. Some records may need to be excluded, and some uses may require permission or other safeguards.

For sensitive or regulated information, involve qualified legal counsel before authorizing access.

You may still have valuable data after removing anything that shouldn't be shared.

3. How will they access and protect it?

Buyers often promise to anonymize data, meaning they remove or change information that could identify people or companies.

Ask how that works before giving anyone access.

For example, Scale AI says it anonymizes records inside the seller's environment before extracting them. It also says sellers approve each data package before release.

Mercor describes an automated system that extracts and de-identifies data. It says AI lab customers don't see the original records.

micro1 says it agrees on the approved records, privacy requirements, and internal permissions before a company participates.

These are the companies' own descriptions of their processes. Your agreement and technical setup need to spell out what will actually happen to your data.

Before connecting any system, ask who can access the original records, where they'll be processed, and what information will leave your company.

You should also know whether you can review the prepared data before it's released and what happens to temporary copies afterward.

Anonymization has limits. Removing names and email addresses doesn't always prevent someone from identifying a person or business. NIST's guidance on de-identification explains why companies need to assess that risk rather than rely on simple masking.

Agree on the handling process and security requirements before sharing records.

4. What are you giving them permission to do?

The amount you're offered matters, but so do the rights the buyer gets.

Can the buyer use the data to train AI models? Can it share the data with other companies? Can it resell the dataset? Can it keep using the information indefinitely?

Find out exactly what rights you're licensing and for how long.

A non-exclusive license lets you retain the ability to license the same data to other buyers, subject to your agreements and legal obligations. An exclusive agreement may prevent you from doing that.

Also clarify whether the license includes future records your business creates.

You can retain ownership of your underlying data while granting a buyer extensive rights to use it. That's why the specific contract terms matter more than simply calling the arrangement a sale or license.

Pay particular attention to AI training.

If a buyer trains a model using your records, deleting the original dataset doesn't necessarily reverse what the model learned. Your contract should explain which uses can continue after the agreement ends and what the buyer must delete or stop using.

If your records contain valuable business methods, pricing information, or customer relationships, consider how the license could affect your competitive position.

5. How and when will you get paid?

Ask for the payment terms before agreeing to prepare or transfer anything.

Some buyers pay after the data has been processed and delivered.

For example, Mercor says it typically pays within two to four weeks of signing, depending on how long extraction and processing take.

Your agreement should be clear about what triggers payment.

Can the buyer reject the prepared dataset? What happens if it doesn't meet their expectations? How long do they have to decide whether to accept it?

Don't leave payment entirely dependent on the buyer's undefined approval.

Agree on what you must deliver, how acceptance will be measured, when payment is due, and how disagreements will be resolved.

For larger deals, consider negotiating payments at agreed milestones.

Also check who pays for data preparation, legal work, security reviews, and any ongoing updates.

Finally, compare the payment with the rights you're granting. A limited license to historical records is a very different commercial arrangement from an exclusive license covering years of future data.

For more context, read How Much Is Company Data Worth?.

What should you say to the buyer?

If you've received an email or LinkedIn message, you can reply with this:

Thanks for reaching out. We're interested in learning more.

Could you share a brief outline of which company would be buying or licensing the data, what records you're looking for, how the data would be used, who would receive it, and the general commercial terms?

We'd also like to understand how you handle anonymization and access to the original records.

Happy to discuss further once we have a better picture of the opportunity.

You can use their response to decide whether a more detailed conversation makes sense.

When should you be concerned?

Be cautious if someone pressures you to move quickly, asks for unrestricted access to your systems, or refuses to identify the company behind the offer.

Other warning signs include unclear payment conditions, broad rights that haven't been explained, and no clear process for protecting sensitive information.

You may be able to resolve some of these concerns by narrowing the data scope or changing the agreement.

But if you can't establish who will use the data, how it will be protected, or whether your company has the right to share it, don't proceed with the transfer.

Already received an offer? Let us review it with you.

DataDeals connects businesses with AI companies interested in licensing their operational data. The service is free for sellers.

If you've been approached, we can help you assess the proposed terms, understand what the buyer is asking for, and decide whether the opportunity is worth pursuing.

Your first 20-minute Data Opportunity Review is free.

Tell us who contacted you, what they're interested in, and the general terms they've proposed.

Send a summary of the offer, not confidential records or raw company data.

Request a Free Data Opportunity Review →