Skip to content
AI Data Payout

Guides

What "de-identified" and "anonymized" mean in plain English

De-identified means the details that point to a person, such as names, email addresses and phone numbers, have been removed or swapped for codes. Anonymized usually means the same thing, but some people use it only when the link to the person can never be rebuilt. Neither word promises that nobody could ever be identified.1,2,3

By Lead analyst (to be named) · reviewed by Reviewing editor (to be named) · published 14 Oct 2026 · updated 14 Oct 2026

The four words, in plain English

These words come up in every data deal, and people use them loosely. The US standards agency NIST says some writers treat de-identification and anonymization as the same, while others keep anonymization for removal that cannot be undone. NIST avoids the word anonymization altogether because of that confusion.2

  1. De-identification: any process that removes the link between data and the people it is about. It is the umbrella word.1
  2. Pseudonymization: swapping names for codes, so all of Jane's messages still show up as Person 14 but not as Jane. Nyne's ads describe this: client names are swapped for codes.1,4
  3. Redaction: simply deleting or blacking out the identifying parts, like a name or an account number.1
  4. Anonymization: used either as another word for de-identification, or for de-identification that cannot be reversed. Ask which one the speaker means.2

Direct and indirect identifiers

Direct identifiers name a person outright: a name, an email address, a phone number, an account number. These are what buyers say they strip out. Cognor Labs, for example, lists names, emails, phone numbers, addresses and account numbers.5

Indirect identifiers are sneakier. A birthday, a ZIP code or a job title does not name anyone by itself, but put a few together and you can often work out who it is. A study found 99.98% of Americans could be correctly re-identified using 15 such traits.6,7

Why email and chat are hard to clean

A spreadsheet has a column marked Name, so it is easy to delete. Free text does not. NIST notes that in written notes, names and addresses may not be clearly marked, useful details can be mistaken for personal ones and removed, and identifying details can survive the cleaning.8

Think of your own inbox: signature blocks, nicknames, "the Hendersons' job on Elm Street", a photo of a driver's licence in an attachment. Each one is a clue a cleaning tool has to catch. That example is ours, but the problem is the one NIST describes.8

There is also a trade-off. NIST says privacy protection improves as cleaning gets more aggressive, but the data becomes less useful. A buyer that cleans too little puts people at risk; one that cleans too much has less to pay for.9

What AI data buyers say they do

Each of these is the buyer's own description. Polyshares says an outside firm scrubs personal data before anything is licensed. Cognor Labs says details are stripped or replaced inside the company's own systems. Mercor says its scrubbing is fully automated. micro1 says it removes personal information and may rewrite some datasets synthetically, so the text no longer ties directly to the original.10,5,11,12,13

The strictest example so far is in court. In the Spirit Airlines sale, an outside firm, Tonic.ai, de-identifies the messages and files a certificate with the court before each transfer.14

Pros and cons

Pros

  • Removing direct identifiers takes out the most obvious risks: names, emails, phone numbers.5
  • Swapping names for codes keeps the data useful while hiding who is who.1,4
  • Outside checks exist: in the Spirit sale, a firm certifies the cleaning before each transfer.14

Cons

  • The words are used inconsistently, so "anonymized" can mean less than it sounds.2
  • Data thought to be anonymous has later been re-identified.3,7
  • Free text like email is the hardest kind of data to clean well.8
  • AI models can sometimes repeat training text, so what slips through can resurface.15

What it means for a business like yours

When a buyer tells you your data will be anonymized, ask three plain questions. Which method: deleting, or swapping for codes? Who does it: the buyer, an outside firm, or your own team? And do you see a sample before anything leaves? micro1, for example, lets sellers review scrubbed samples first.16,10

Then make it easier for the cleaner. Leave out the folders most full of personal details, like HR files and customer account records, before the data goes anywhere. Nyne's ads say any folder can be excluded.17,18

Should your business get in now?

Many businesses like yours are already licensing their records to these firms. micro1 has committed $1B over 12 months to buying company data. At least 18 brokers started advertising for business data between September and mid-October 2026. SimpleClosure has done over 150 data deals, and Google bid $10M for Spirit Airlines' archive, though the court has not ruled on that sale yet.19,20,21,22

Buyers also publish deals by industry. Polyshares, for example, lists a $110,000 license for a marketing agency. That is its own claim, like every payout a buyer or broker publishes about itself.23

Our view, not a promise: the money will not be around forever. AI labs are racing to buy now, budgets are concentrated in today's build-out, and as more businesses sell, prices tend to come down. There is no deadline, but waiting has a cost.19,24

It is not for everyone

Think twice if any of these sound like you:

  • Most of your records are client files you hold in confidence.25
  • Your team would feel their messages were sold out from under them.26
  • Your data is regulated or sensitive, which buyers ask about up front.27
  • Your archive is small: most buyers want 20-30+ staff and a few years of records.28,29,30,31
  • You cannot spare the weeks a deal takes.32,33,34
  • You are not comfortable losing control of how a licensed copy is used.35

Next step

Take the quiz to see which firms fit your business. Then talk to the brokerages: they can answer the questions unique to your business. Legal questions go to the licensing companies.

Take the quizWho the brokerages are

We feature companies that pay us a referral fee. Our scores and reviews don't change for it. How we earn.

FAQ

Questions, answered with sources

  • The details that point to a person, such as names, emails and phone numbers, have been removed or swapped for codes. It is the umbrella word for this kind of cleaning.1

  • Usually the same as de-identified. Some people use it only for removal that cannot be reversed, which is why NIST avoids the word.2

  • Data where names are swapped for codes, so records about the same person still link together but do not show who it is.1

  • Sometimes. Data thought to be anonymous has been re-identified, and one study found 99.98% of Americans could be identified from 15 traits.3,7

  • In free text, names and addresses are not clearly marked, and identifying details can survive cleaning. NIST describes the same problem in written notes.8

  • It varies. Polyshares uses an outside firm, Cognor Labs says it works inside your systems, Mercor automates it, and in the Spirit sale an outside firm certifies it to the court.10,5,11,14

  • We do not answer legal questions. Ask the licensing company: their team reviews this with you before anything is signed.

Before you sign

Legal questions go to the licensing company

We do not answer legal questions. Ask the licensing company before you sign: their team reviews these with you.

Questions for this kind of business

  • Which de-identification method will you use, and will an outside party certify it?
  • How do CCPA, GDPR, UK GDPR and state privacy laws apply, and is de-identified data still personal information?

Questions for any buyer

  • Is the license exclusive, and for how long?
  • Can you sublicense the data, and to whom?
  • Which AI labs will receive it, and will you tell us?
  • How long do you keep it, and how is it deleted?
  • What is your liability cap, and who indemnifies whom?
  • Will the contract ban re-identification of people in the data?
  • Who removes personal data, by what method, and do we approve it first?

How this page was made

Every figure comes from a published source listed below, each with the date we checked it. Buyer statements are the buyer's own claims and are marked so in the source list. Firm counts come from the US Census Bureau's Statistics of US Businesses. Anything marked as our estimate is ours, not a buyer quote.

Drafted with AI assistance from our research files, then checked against each source. Not yet reviewed by a person, so this page is not indexed by search engines. Last updated 14 Oct 2026.

We feature companies that pay us a referral fee. Our scores and reviews don't change for it. See how we earn and how we rank. This site does not give legal, tax or financial advice.

Sources (35)

  1. 1.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  2. 2.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  3. 3.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  4. 4.Meta Ad Library, Nyne.ai. Accessed 14 Oct 2026. Company's own claim.
  5. 5.Cognor Labs website. Accessed 14 Oct 2026. Company's own claim.
  6. 6.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  7. 7.Rocher, Hendrickx and de Montjoye, Nature Communications, published 23 Jul 2019. Accessed 14 Oct 2026.
  8. 8.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  9. 9.NIST IR 8053, De-Identification of Personal Information, published 1 Oct 2015. Accessed 14 Oct 2026.
  10. 10.Polyshares website. Accessed 14 Oct 2026. Company's own claim.
  11. 11.Mercor Data page. Accessed 14 Oct 2026. Company's own claim.
  12. 12.beri.net, published 10 Oct 2026. Accessed 14 Oct 2026.
  13. 13.micro1 data partnerships page. Accessed 14 Oct 2026. Company's own claim.
  14. 14.The Next Web, published 9 Oct 2026. Accessed 14 Oct 2026.
  15. 15.arXiv (Carlini et al.), published 14 Dec 2020. Accessed 14 Oct 2026.
  16. 16.beri.net, published 10 Oct 2026. Accessed 14 Oct 2026.
  17. 17.Meta Ad Library, Nyne.ai. Accessed 14 Oct 2026. Company's own claim.
  18. 18.SimpleClosure press release (Business Wire), published 3 Sep 2026. Accessed 14 Oct 2026. Company's own claim.
  19. 19.The Next Web, published 9 Oct 2026. Accessed 14 Oct 2026.
  20. 20.Meta Ad Library (our review). Accessed 14 Oct 2026.
  21. 21.Forbes, published 16 Apr 2026. Accessed 14 Oct 2026.
  22. 22.Bloomberg Law, published 17 Aug 2026. Accessed 14 Oct 2026.
  23. 23.Polyshares track record page. Accessed 14 Oct 2026. Company's own claim.
  24. 24.TechCrunch, published 16 Sep 2025. Accessed 14 Oct 2026.
  25. 25.Polyshares website. Accessed 14 Oct 2026. Company's own claim.
  26. 26.Slashdot discussion, published 18 Apr 2026. Accessed 14 Oct 2026.
  27. 27.micro1 data payout estimator. Accessed 14 Oct 2026. Company's own claim.
  28. 28.micro1 company referral page. Accessed 14 Oct 2026. Company's own claim.
  29. 29.Nyne sell-company-data page. Accessed 14 Oct 2026. Company's own claim.
  30. 30.Cognor Labs partners page. Accessed 14 Oct 2026. Company's own claim.
  31. 31.Return My Time Sell My Data page. Accessed 14 Oct 2026. Company's own claim.
  32. 32.RemoWork, published 29 Jun 2026. Accessed 14 Oct 2026.
  33. 33.Your Data Value website. Accessed 14 Oct 2026. Company's own claim.
  34. 34.Cognor Labs manufacturing page. Accessed 14 Oct 2026. Company's own claim.
  35. 35.micro1 Data License Master Terms, published 7 Oct 2026. Accessed 14 Oct 2026. Company's own claim.