// Field notes / Data licensing
Is it legal to license your company's emails, CRM and calls for AI training?
By Rob Eason, founderPublished
// The starting point
Holding the records isn't the right to sell them
Most owners assume that records the company created belong to the company to sell. The lawyers writing about this market start from the opposite end. Frankfurt Kurnit Klein + Selz put it in one line: a company that holds data does not necessarily have the right to sell it. GC AI, writing about customer data, says having the data isn't the same as having the right to train on it.
The FTC's staff said the same about the AI side in January 2024: there is no AI exemption from the laws on the books.
So the question isn't whether licensing company data is legal in general. It's whether you may license these particular records. The answer comes from four places, and each one can rule records out.
- Your promises: the privacy policy and terms in effect when each record was collected.
- Your contracts: customer agreements, NDAs, and any data you licensed from someone else.
- The law on certain kinds of data: state privacy laws, health and financial records, children's data, biometrics, recorded calls.
- The tools that hold the records: the terms of the software you'd export from.
// Your promises
The privacy policy you had then still applies
The rule both law firms state is that you must honor the privacy statement in effect when the data was collected. A policy you update today doesn't reach back over records you gathered under an older one.
That covers more than the policy page. The FTC's staff wrote that companies must keep their commitments to customers regardless of how or where the commitment was made. A promise in a sales email, a contract or an FAQ counts.
Changing the rules after the fact has its own risk. The FTC warned in February 2024 that quietly amending terms or a privacy policy to allow AI training may be unfair or deceptive, and its January 2024 post points to the need for clear and conspicuous notice and affirmative express consent before data is used in new ways.
Buyers have reason to care too. The FTC's staff noted that it has required businesses that obtained consumer data unlawfully to delete products built with it. A careful buyer will ask what you promised before it trains on anything.
// Your contracts
Customer agreements, NDAs and data you licensed in
LumaLex Law notes that customer contracts often limit how you can use their data. If you provide a service, read your own standard agreement first. GC AI gives a common clause as an example: the provider may process customer content solely to provide the services to the customer. A clause like that leaves no room to license the content to someone else.
Email and chat are the hardest records to clear. Frankfurt Kurnit points out that they hold other companies' confidential information, often covered by NDAs and commercial agreements. A vendor's pricing, a partner's plans and a customer's numbers can all sit in one inbox.
Records you licensed from someone else come with their own terms. Purdue Global Law School notes that data licenses may not permit AI uses if those uses aren't spelled out in the agreement. A purchased list, a data feed or an industry database is usually not yours to pass on.
// The law on certain data
State privacy laws and regulated records
Some records carry rules that apply whatever your contracts say. These are the ones the law firms flag most often.
- California residents' information. Frankfurt Kurnit reads a transfer of workplace data for money as a sale of personal information under California law, and the California Attorney General lists the right to opt out of the sale or sharing of personal information among the rights that law gives.
- Sensitive data in other states. LumaLex notes that Maryland and New Jersey prohibit the sale of sensitive data.
- Patient records. Under HIPAA, LumaLex writes, licensing patient data typically needs a valid written authorization from each affected patient.
- Financial, children's and biometric data. Frankfurt Kurnit lists GLBA, children's privacy laws and biometric laws among the rules to check.
- Recorded calls. The same article lists call recording consent laws. If your records include call recordings, check what callers agreed to when they were recorded.
// Your people and your tools
Employee messages and platform terms
Your staff wrote most of the email and chat a buyer wants. Their messages carry their own protections. LumaLex notes that messages about pay, scheduling and working conditions may be protected concerted activity, and Frankfurt Kurnit says the same of employees objecting together under Section 7 of the National Labor Relations Act. A union contract can add limits of its own.
Then check the software. Both firms point out that Slack's API terms prohibit third-party apps from using API data to train large language models. Other tools have terms of their own on exports and use, so read each one before you plan what to pull.
// Before you send anything
A sample is already a disclosure
Buyers often ask for a sample early. Both law firms say not to send one before the legal review is done, because a sample is itself a disclosure.
Removing names doesn't settle the question either. GC AI says calling the data deidentified is not enough, and the FTC's staff wrote in 2024 that data is only anonymous when it can never be associated back to a person. Frankfurt Kurnit advises leaving high-risk data out entirely rather than relying on de-identification.
And think of the license as permanent, whatever its term says. Purdue Global Law School notes that removing data from a trained model can be virtually impossible. Deletion clauses cover the raw files, not what a model learned from them.
// For your lawyer
What to bring to the first meeting
None of this is legal advice. A privacy lawyer decides what you may license, and the meeting goes faster with these in hand.
- Every version of your privacy policy and customer terms, with the dates each was live.
- Your standard customer agreement and your largest customers' signed contracts.
- A list of NDAs and confidentiality clauses with partners and vendors.
- A list of systems that hold records, with the years each covers and the kinds of data inside.
- Any data you bought or licensed from others, with its terms.
- Union contracts or employee policies that mention monitoring or company communications.
- The buyer's draft agreement, with its permitted uses. LumaLex suggests narrowly defined uses, such as model training only, no resale, no re-identification and no competing products.
// Questions
What owners ask about the law
Is it legal to sell customer data to AI companies?
Can I sell my company's emails to train AI?
Do I need customer consent to license data for AI training?
Does de-identifying the data make it legal to sell?

Rob Eason, founder
// Book a call with Rob
Sorting out which of your records could be licensed?
A 30-minute call with Rob. He isn't a lawyer, and the call isn't legal advice. He'll help you list which systems hold what, sort the questions that belong with your lawyer, and look at what the same records could do inside your own company.
- Co-owned a 60-employee financial services firm that served more than 7,000 customers and has a 5-star rating across more than 600 customer reviews, and sold his share to his partner.
- Ran GOSO Technologies, a Boston solar company that Forbes Home named one of the top 10 solar companies in Boston, rated 5/5 across 24 Google reviews.
- Spoke on the construction panel at Build Connect 2026 in Charlotte, sponsored by Choate Construction.
- HighLevel Certified Admin with the second-most reviews in HighLevel's Certified Admin Directory worldwide: 5/5 across 83 reviews.
- Won HighLevel's Top New Subaccounts via AI Employee award in its 2025 Summer of AI Awards.
- 4.3/5 from 6 Google reviews of Massmarket AI.
- Every directory and Google review, critical ones included, is shown word for word at massmarket.ai/reviews.
- Takes every proof-of-concept call himself.
