The party that bears the risk or owns the file can license claims data: a carrier, including one in run-off, an MGA for its own programs, a captive or a self-insured group. A TPA or independent adjuster usually handles files for a client and needs that client's written consent. Buyers want the claim from first notice to payout, with reserve changes and adjuster notes. No buyer publishes a price for claims files; Mercor lists claims among the finance and insurance data it wants.
- Who may sign decides more claims deals than price does: the file owner signs, the administrator usually cannot alone.
- Reserve history linked to the final payout is the part buyers value most.
- Bodily-injury and workers' compensation files carry health information: de-identify them properly or leave them out.
- No claims-specific price has been published; general company ranges are ceilings, not offers.
Who holds claims files, and who can sign a license
Claims files are held by several kinds of company, and holding a file is not the same as owning it. The table is the usual position; your service agreements and program agreements decide the real one.
This page is general information, not legal advice: it quotes each rule with its source, and your counsel decides how it applies to your company.
| Holder | Usual position on the file | Can it sign alone? |
|---|---|---|
| Carrier, including a book in run-off | Owns its claim files | Yes, subject to privacy law and reinsurance terms |
| MGA with its own programs | Depends on the program agreement with the carrier | Only if the agreement gives it the data rights |
| Captive or self-insured group | Owns its own claims history | Yes, through its board |
| Third-party administrator (TPA) | Handles files for clients under a service agreement | Usually not; needs written client consent |
| Independent adjusting firm | Works files assigned by carriers | Usually not; its own methods and training material are its own |
What AI buyers want from a claims file
Buyers building claims agents want the whole life of a claim and how it ended. Mercor lists "Finance & Insurance Statements, underwriting files, claims" among the data it seeks (Mercor).
| Part of the file | What it shows | Why it matters |
|---|---|---|
| First notice of loss | What was reported, when and how | The starting facts |
| Adjuster notes and diary | Investigation, coverage questions, decisions | The reasoning behind each step |
| Reserve history | Each reserve change and its date | Shows how the expected cost moved as facts came in |
| Photos and estimates | Damage and the estimate written from it | Links evidence to the amount |
| Payments and closing | What was paid, to whom (coded) and when | The outcome that labels the rest |
| Litigation and subrogation flags | Whether the claim went to suit or recovery | Marks the hard cases |
Underwriting submissions and quote-or-decline decisions are a separate data set; see licensing insurance underwriting files.
What claims data is worth: the honest evidence
No buyer publishes a price for claims files, and no closed claims-data deal has been published with a source. The figures that exist are general company ranges.
| Evidence | What it says | Grade |
|---|---|---|
| micro1 company data program | Payments "typically ranging from $100K to over $3M" (micro1) | A buyer's published range, all industries |
| Mercor data program | Lists "Finance & Insurance Statements, underwriting files, claims" as target data (Mercor) | Buyer statement of demand; no claims price |
These ranges are ceilings a buyer advertises across all industries, not offers and not a price for claims files. What moves value is years of closed claims with reserves and payouts intact.
Legal limits: privacy law and health information
Federal financial privacy law limits disclosure of personal information. Under the Gramm-Leach-Bliley Act, "A financial institution may not disclose nonpublic personal information to a nonaffiliated third party" unless the consumer has been told and given a chance to opt out (15 U.S.C. 6802). State insurance privacy laws add their own rules. In practice, claimant names, addresses, policy numbers and payment details are removed or coded before anything is licensed.
Bodily-injury and workers' compensation files often contain medical records, and they are the most restricted part of a claims file. HIPAA rarely governs them: its definition of a health plan excludes "excepted benefits" (45 CFR 160.103), which include "Liability insurance, including general liability insurance and automobile liability insurance" and "Workers’ compensation or similar insurance" (42 U.S.C. 300gg-91). The binding rules are state insurance privacy laws. In states that adopted the NAIC's privacy regulation (Model 672), an insurer "shall not disclose nonpublic personal health information about a consumer or customer unless an authorization is obtained from the consumer or customer whose nonpublic personal health information is sought to be disclosed" (New York's version: 11 NYCRR 420.17), and a consumer includes an individual who "is a claimant under an insurance policy issued by the licensee" (11 NYCRR 420.3). Licensing to an AI developer is not one of the listed insurance functions, such as claims administration and reinsurance, that need no authorization. Leave medical content out unless counsel confirms a lawful route, and de-identify the rest. Where HIPAA does apply, expert de-identification requires a finding that "the risk is very small" of re-identification (45 CFR 164.514). See how to de-identify company data.
Records checklist for a claims-file holder
The sector records checklist (CSV) has these rows for insurance and the other four sectors.
| Item | What to write down |
|---|---|
| Claims system | Guidewire ClaimCenter, Duck Creek, Origami Risk, Ventiv or similar; years in each |
| Lines of business | Property, auto physical damage, general liability, workers' compensation and so on |
| Volume | Closed claims by line and year |
| Completeness | Whether reserve history, notes, photos and payments survive for closed claims |
| Ownership | For TPAs and MGAs: which clients or carriers own which files, and their data clauses |
| Exclusions | Medical records, open litigation, files under a client prohibition |
What a one-page datasheet looks like (fictional example)
Invented to show the level of detail; it describes no real company.
| Field | Fictional example |
|---|---|
| Company | Program MGA writing small commercial property and general liability, 60 staff |
| Years | Claims system 2015 to 2026 |
| Volume | About 48,000 closed claims with reserve history and payments |
| Rights | Program agreement gives the MGA use of program data; carrier consent letter on file |
| Exclusions | Bodily-injury medical records removed; claimant identities coded |
| Signer | CEO, after review by general counsel |
What happens next: three steps
A first conversation needs no files. You can stop after any step.
- Describe your records: answer six questions, with no files and no system access. Origin then checks which files you own, which belong to clients or carriers, and what consent is needed.
- Datasheet and sample: Origin turns your answers into a one-page datasheet like the example above, and you approve a small de-identified sample before anything leaves the company.
- Buyers and license: Origin shows the datasheet to the AI developers and data programs that have asked for claims data, with their published terms. You approve scope, use, price and delivery, and sign the license yourself.
- Origin does not charge data owners; a buyer may pay Origin when a deal closes (how Origin is paid).
Resources for claims managers
Guides for the people who keep these records:
- Loss run report: what it shows and how to read one
- Can you sell insurance claims data to AI companies?
- Licensing insurance underwriting files
- De-identification before licensing
Hold years of closed claims files? Describe them in six questions, no files needed, and see what your data is worth.
Origin Data Partners works with companies that own data and helps get it bought. We are independent of the services on this page, and a buyer may pay us when a deal closes (how Origin is paid). See what your data is worth.
Common questions
Can a third-party administrator sell the claims files it handles to AI companies?
Usually not on its own. A TPA normally handles files for clients under a service agreement, so the client decides what can be licensed and a TPA needs written client consent. The TPA's own procedures, training material and anonymous process statistics may be its own to license.
Who owns the claims files that a TPA handles for its clients?
In most service agreements, the client: the carrier or self-insured employer that bears the risk. Read the data and confidentiality clauses of each agreement; they decide ownership and use, not who holds the files.
Can an insurer in run-off license its claims history to AI developers?
A carrier in run-off owns its claims files, so it can license them subject to privacy law, its reinsurance agreements and any regulator conditions. Remove claimant identities and medical content first.
What about bodily-injury and workers' compensation medical records in claims files?
Treat them as the most restricted part. State insurance privacy rules based on NAIC Model 672 require the person's authorization before an insurer discloses health information, and HIPAA rarely applies to property-casualty or workers' compensation claims. Most sellers leave medical content out and keep the rest of the file.
What is an insurance claims file worth to an AI company?
No buyer publishes a claims-specific price. General company programs publish ranges such as micro1's $100K to over $3M across all industries, and Mercor lists claims among the data it wants. Those are ceilings, not offers.
Which claims systems export cleanly enough for a data license?
Any modern claims system can export claims, reserves, payments and notes; what matters is whether history survived system changes. Check that reserve changes, notes and photos are still attached to closed claims before describing the archive.
What does Origin charge a claims-file holder?
Nothing. Origin does not charge data owners; a buyer may pay Origin when a deal closes. You approve every step and sign the license yourself.
What is claims data?
Claims data is the record an insurer or claims administrator keeps for each claim: first notice of loss, the coverage decision, adjuster notes, reserves and how they changed, payments, recoveries and the outcome. Health insurers' billing claims are a different market and are not what this page covers.
What is insurance run-off, and who holds the claim files?
A run-off book is a block of policies an insurer no longer writes or renews; its open claims are still handled until they close. The insurer, a run-off company that acquired the book or a TPA handles them, and the owner of the book usually controls the files.
Is healthcare claims data the same as an insurance claims file?
No. Healthcare claims data is the billing record between a provider and a health plan: the claim sent, the payer's decision and the remittance that explains it, such as an 835 electronic remittance advice. A property, casualty or workers' compensation claims file follows one loss from first notice to payout, with coverage decisions, adjuster notes and reserve changes. Billing and revenue-cycle records are covered in the healthcare billing and prior authorization guide; for how open and closed claims appear on a carrier report, see reading a loss run.
Sources and further reading
- Mercor: data monetization
- micro1: Company Data Partnerships
- 15 U.S.C. 6802: obligations with respect to disclosures of personal information (Gramm-Leach-Bliley Act)
- 45 CFR 160.103: definitions (health plan excludes excepted benefits)
- 42 U.S.C. 300gg-91(c)(1): excepted benefits (liability, workers' compensation, auto medical payment insurance)
- N.Y. Comp. Codes R. & Regs. tit. 11, 420.17: authorization required for disclosure of nonpublic personal health information (New York adoption of NAIC Model 672)
- N.Y. Comp. Codes R. & Regs. tit. 11, 420.3: definitions (consumer includes a claimant)
- 45 CFR 164.514: de-identification of protected health information