Skip to content
Company Data Advisors

Blog

Can you license records that include employee and customer data?

· Updated · Company Data Advisors

Often yes, but not as stored. What the Spirit Airlines case, one law firm's checklist and five buyers' pages say about employee, customer and vendor data.

Our reading: usually yes, but not as the records sit in your systems. Email, chat and tickets are full of employees', customers' and vendors' information, so buyers say they remove or mask it before it reaches their models, and some categories are better left out altogether. Where the removal happens varies by buyer, and the company that holds the records is the party that should be able to show it had the right to license them. This guide pulls together the only court record we found that deals with the question, a lawyer's published checklist and what buyers say they do.

This is a reading of public sources, not legal advice. The answer depends on your privacy notices, your contracts, your state and your industry, so have counsel review before you share anything, including a sample.

Who is responsible for the personal information in your records?

The company that collected and holds the records is responsible for deciding whether it may license them. Daniel Goldberg of Frankfurt Kurnit, writing in Legal 500 on October 6, 2026, says a company that holds data "does not necessarily have the right to sell it" and that "most companies cannot sell everything a buyer requests." Slack's User Terms put a similar duty on the customer: as between Slack and the customer, it is solely the customer's responsibility to obtain the rights, permissions or consents from its users that lawful use of the data requires, as we explain in our guide to Slack, Google and Microsoft terms.

Goldberg's concerns include four places a company's obligations come from:

  • Your own privacy statements. Companies "must comply with the privacy statements in effect when the data was collected." He notes that most privacy policies, employee notices and handbooks "do not contemplate selling workplace communications for AI training."
  • State privacy law. For California, he writes that a transfer of workplace data for money or other valuable consideration is a sale of personal information unless the data meets the CCPA's definition of de-identified. That definition requires technical measures, a public commitment not to re-identify and contractual obligations on recipients.
  • Labor law. Employee messages about pay, scheduling and working conditions "may be protected concerted activity" under Section 7 of the National Labor Relations Act, whether or not employees are unionized. Where employees are unionized, the collective bargaining agreement may restrict a sale.
  • Contracts with other companies. Email and chat "contain other companies' confidential information, often covered by NDAs and commercial agreements."

Emory law professor Ifeoma Ajunwa told Business Insider in October 2026 that the lack of federal privacy laws leaves few legal barriers to collecting data from employees for AI training. She called it "an ethically gray area" whether employees' consent "actually extend[s] to the sale of the data." In our view, that gap is the reason a company should decide its own rules rather than wait for a rule to tell it.

What does the Spirit Airlines case show?

The Spirit Airlines bankruptcy is the only public court record we found of a proposed sale of a company's internal records, including employee data, for AI training. In an August 2026 auction, Google was selected as the winning bidder for Spirit's de-identified enterprise data, subject to court approval, according to the court-appointed privacy ombudsman's supplemental report dated October 5, 2026 (S.D.N.Y. Case No. 25-11897, Doc. 1684). The report lists what was originally planned for the sale: email and calendar data, document repositories, chat messages, wikis and several core business systems.

The report describes four protections the parties agreed to:

  1. Whole systems left out. Spirit and Google agreed to exclude the passenger transaction, refund, in-flight purchase and Wi-Fi sales systems (the list was still subject to change), so that data would not go to Google "in any form (deidentified or otherwise)."
  2. An independent de-identifier. Spirit hired a third-party agent, Tonic.ai, to de-identify unstructured data such as corporate email before transfer.
  3. A named standard. The parties agreed to follow the CCPA's de-identification standard: reasonable measures so data cannot be associated with a person, a public commitment by the purchaser not to re-identify, and a contractual ban on passing the data on unless the next party makes the same commitment.
  4. Certification as a closing condition. The agent must file an affidavit with the court before every transfer saying the data has been de-identified.

The report also contains a limit that matters to any employer. A footnote says the ombudsman "did not investigate or reach any conclusion about whether the data offered for sale or the deidentification process to be followed presents any risks to affected employees or third parties." The ombudsman's report concerns consumers' personal information under the bankruptcy code, and employees fell outside her review.

Employees and others noticed. Spirit's flight attendants' union and a software vendor that says some of the data belongs to it filed objections, according to Legal 500, and Business Insider reported that Senator Elizabeth Warren and more than 100 other members of Congress wrote to Google and Spirit about former employees' privacy. After a hearing adjourned from November 6 (the debtors' notice of adjournment is dated November 4, 2026), Judge Sean Lane approved the sale on November 17, according to Bloomberg Law, saying a recent agreement with the flight attendant and pilot unions provides appropriate protections. The revised deal, as WIRED reported on November 12, lets Google walk away if cleaning the data shrinks it by more than 8 percent; our post on what happens if a buyer rejects part of your data explains why that matters to any owner. We have not yet read the signed order, so check the docket for its final terms.

Our view: Spirit shows what a careful process looks like on paper, and what it leaves open. The proposed sale excludes whole systems, uses an independent agent, names a standard and requires certification, yet the sale as designed in October was still reported to include payroll and tax records, according to the Miami Herald (October 13), and the ombudsman did not assess employee risk. Opposition from the union was reported to be softening by mid-October, and the court approved the sale in November after a revised agreement with the unions. A company that is operating, with current staff and current customers, has more to lose from that reaction than a company that has closed.

Which records should you consider leaving out?

Goldberg advises excluding high-risk data "entirely rather than relying on de-identification." His list of what to consider excluding, with his reasons, is below.

Records What Goldberg says
Privileged material A sale "may waive privilege over legal communications"
HR and personnel files On his exclusion list
Payroll and tax records On his exclusion list
Health and benefits information Internal data may include information subject to HIPAA
Union-related communications Employee communications about pay and working conditions may be protected under the NLRA
Customer and partner confidential information Often covered by NDAs and commercial agreements
Licensed data and vendor-generated data May come with restrictions or competing ownership claims
Data from regulated systems May be subject to laws such as HIPAA and GLBA

Goldberg adds that de-identification "addresses whether a record can be linked to an individual, not whether its content is confidential." In his words, "a disciplinary thread or pay complaint remains sensitive without a name attached." He also warns that buyers want records to stay linked to each other, which is what makes the data valuable and what makes it "possible to draw conclusions about individuals or small groups." Our guide to how to sell company data to AI lists what buyers want.

How do buyers say they handle personal information?

Buyers' published processes differ on who removes personal information, where it happens and what stays with you. The five below describe it on their own program pages. Only Scale says the removal happens inside your environment before extraction; the others describe it after export or delivery.

Program Who removes personal data What else it says
Scale AI Scale, "inside your environment, before any data is extracted" Per-package approval gates and a full audit trail
Turing Turing, in a dedicated air-gapped environment, working from a copy you provide "Personal data and secrets are removed before anything reaches a model"
Collinear AI Collinear, after export; its page lists personal data as something "we remove either way" Names are replaced with consistent pseudonyms; you can hold back whatever you prefer not to send
The General Data Company (industrial companies) Its own model finds and replaces identifiers, a person checks the result, then it deletes what links new names to real ones A confidentiality and data-processing agreement comes first; current designs and prices remain yours
SourceX SourceX or its processing partner, after the data reaches SourceX; its process page says this applies where the signed deal confirmation requires it Its fit check asks for no customer or employee records; it says its agreement records that re-identification may still be possible

Our pages on Scale AI, Turing, Collinear AI, The General Data Company and SourceX give each program's process and sources. A buyer that does the removal for you is not the same as a buyer that takes your sign-off on a sample, and neither is the same as an independent third party you choose. In Spirit, an independent agent was written into the sale agreement. Goldberg advises that the company select or jointly select the de-identification provider, approve the protocol and require certification "to the company's satisfaction."

The SourceX row deserves a second read. SourceX says its agreement records that no method removes every possibility of re-identification, which is one more reason to decide what stays out before the buyer's process starts. For what that clause means for your license, see what you're agreeing to.

What about customers' and vendors' information?

Customers' and vendors' information is covered by your contracts, not only by privacy law. Goldberg says to review "customer, partner, vendor, and data license agreements," to identify data you hold "as a service provider or processor," and to check the terms of the systems the data comes from. In Spirit, a software vendor objected that records its platform generated belong to the vendor, and the vendor's chief executive, Doug Kreuzkamp, told the Miami Herald that Spirit's representatives had "agreed they have data in their possession that belongs to Springshot," while disagreements remained over the full scope.

Goldberg also warns that a sale may not stay quiet: California's AB 2013 requires generative AI developers to disclose publicly, in high-level terms, the sources of their training data and whether it was purchased or licensed. Customers and vendors may read about your deal even if you never announce it.

The same logic applies in reverse. Goldberg notes that vendors holding your data, particularly HR and collaboration vendors, may sell it when they are acquired or shut down, and that contract language on sales, transfers and de-identified or aggregated data is easier to negotiate before an offer arrives.

What should you do before you share anything?

Do the work in this order, because each step limits the next:

  1. Inventory the records. List each system, record type, volume and start date, and mark each as included or excluded, as Goldberg suggests.
  2. Read your own promises. Check privacy notices, employee handbooks and customer terms for what they say about sharing and AI use.
  3. Check your contracts. Look for NDAs, license terms and data processing agreements that restrict reuse, and your software providers' terms for the extraction route.
  4. Decide the exclusions. Use the table above as a starting list and add anything your own contracts or industry rules protect.
  5. Ask each buyer who removes personal data, where, and who certifies it. Get the answer in writing, along with what happens if a record slips through.
  6. Ask whether anyone needs to be told. That includes employees, customers, a union and the vendors whose systems the data comes from. Counsel should answer this one.
  7. Hold the sample. Goldberg cautions against sharing sample data before this review, "since a sample itself could be a disclosure."
  8. Write down what you decided and why. If the company is later acquired, a prior data sale will come up in diligence.

Where we come in

Company Data Advisors is an independent advisory firm. We track what each buyer publishes about how it handles personal information and which records it excludes, and we read the terms with the owner, including who removes personal data and what the contract says if something is missed. Nothing about your company is shared until you approve a specific, named buyer. Participating buyers may pay us referral fees, and some programs in the table above are among them; we tell you which before any introduction. Owners pay us nothing, and you can apply to any buyer directly. We do not give legal advice, so the questions above about notice, consent and labor law belong with your counsel.

If you are weighing an offer, start with steps 1 to 3 above before you fill in any intake form. When you are ready to compare programs, start with the screener. For the full path from first form to payment, see how to license your company's data to AI companies, and for what to check about the buyer itself, see how to check an AI data buyer is legitimate.

Programs in this post

See which buyers fit your company

Eight quick questions against every program in our buyer index. Participating buyers may pay us referral fees.